Re: [PATCH v3 3/5] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Yury Norov]

On Mon, Jul 17, 2023 at 01:37:06PM +0200, Alexander Potapenko wrote:
> The config implements the EA0 algorithm suggested by Evgenii Stepanov
> to compress the memory tags for ARM MTE during swapping.
> 
> The algorithm is based on RLE and specifically targets 128-byte buffers
> of tags corresponding to a single page. In the common case a buffer
> can be compressed into 63 bits, making it possible to store it without
> additional memory allocation.
> 
> Suggested-by: Evgenii Stepanov <eugenis@xxxxxxxxxx>
> Signed-off-by: Alexander Potapenko <glider@xxxxxxxxxx>
> 
> ---
>  v3:
>   - Addressed comments by Andy Shevchenko:
>    - use bitmap_{set,get}_value() writte by Syed Nayyar Waris
>    - switched to unsigned long everywhere (fewer casts)
>    - simplified the code, removed redundant checks
>    - dropped ea0_compress_inline()
>  - added bit size constants and helpers to access the bitmap
>  - explicitly initialize all compressed sizes in ea0_compress_to_buf()
>  - initialize all handle bits
> 
>  v2:
>   - as suggested by Yuri Norov, switched from struct bitq (which is
>     not needed anymore) to <linux/bitmap.h>
>   - add missing symbol exports
> ---
>  arch/arm64/Kconfig               |  10 +
>  arch/arm64/include/asm/mtecomp.h |  60 +++++
>  arch/arm64/mm/Makefile           |   1 +
>  arch/arm64/mm/mtecomp.c          | 406 +++++++++++++++++++++++++++++++
>  4 files changed, 477 insertions(+)
>  create mode 100644 arch/arm64/include/asm/mtecomp.h
>  create mode 100644 arch/arm64/mm/mtecomp.c
> 
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index a2511b30d0f67..52cdc7603cf7c 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -2093,6 +2093,16 @@ config ARM64_EPAN
>  	  if the cpu does not implement the feature.
>  endmenu # "ARMv8.7 architectural features"
>  
> +config ARM64_MTE_COMP
> +	bool "Tag compression for ARM64 MTE"
> +	default y
> +	depends on ARM64_MTE
> +	help
> +	  Enable tag compression support for ARM64 MTE.
> +
> +	  128-byte tag buffers corresponding to 4K pages can be compressed using
> +	  the EA0 algorithm to save heap memory.
> +
>  config ARM64_SVE
>  	bool "ARM Scalable Vector Extension support"
>  	default y
> diff --git a/arch/arm64/include/asm/mtecomp.h b/arch/arm64/include/asm/mtecomp.h
> new file mode 100644
> index 0000000000000..0c444c0d4ac04
> --- /dev/null
> +++ b/arch/arm64/include/asm/mtecomp.h
> @@ -0,0 +1,60 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef __ASM_MTECOMP_H
> +#define __ASM_MTECOMP_H
> +
> +#include <linux/types.h>
> +
> +/*
> + * ea0_compress() - compress the given tag array.
> + * @tags: 128-byte array to read the tags from.
> + *
> + * Compresses the tags and returns a 64-bit opaque handle pointing to the
> + * tag storage. May allocate memory, which is freed by @ea0_release_handle().
> + */
> +unsigned long ea0_compress(u8 *tags);
> +
> +/*
> + * ea0_decompress() - decompress the tag array addressed by the handle.
> + * @handle: handle returned by @ea0_decompress()
> + * @tags: 128-byte array to write the tags to.
> + *
> + * Reads the compressed data and writes it into the user-supplied tag array.
> + * Returns true on success, false on error.
> + */
> +bool ea0_decompress(unsigned long handle, u8 *tags);
> +
> +/*
> + * ea0_release_handle() - release the handle returned by ea0_compress().
> + * @handle: handle returned by ea0_compress().
> + */
> +void ea0_release_handle(unsigned long handle);
> +
> +/* Functions below are exported for testing purposes. */

Then declare them in a separate local header or simply in the test, but
please not in a public header.

> +
> +/*
> + * ea0_storage_size() - calculate the memory occupied by compressed tags.
> + * @handle: storage handle returned by ea0_compress.
> + */
> +int ea0_storage_size(unsigned long handle);
> +
> +/*
> + * ea0_tags_to_ranges() - break @tags into arrays of tag ranges.
> + * @tags: 128-byte array containing 256 MTE tags.
> + * @out_tags: u8 array to store the tag of every range.
> + * @out_sizes: u16 array to store the size of every range.
> + * @out_len: length of @out_tags and @out_sizes (output parameter, initially
> + *           equal to lengths of out_tags[] and out_sizes[]).
> + */
> +void ea0_tags_to_ranges(u8 *tags, u8 *out_tags, short *out_sizes, int *out_len);
> +
> +/*
> + * ea0_ranges_to_tags() - fill @tags using given tag ranges.
> + * @r_tags: u8[256] containing the tag of every range.
> + * @r_sizes: u16[256] containing the size of every range.
> + * @r_len: length of @r_tags and @r_sizes.
> + * @tags: 128-byte array to write the tags to.
> + */
> +void ea0_ranges_to_tags(u8 *r_tags, short *r_sizes, int r_len, u8 *tags);
> +
> +#endif // __ASM_MTECOMP_H
> diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
> index dbd1bc95967d0..46778f6dd83c2 100644
> --- a/arch/arm64/mm/Makefile
> +++ b/arch/arm64/mm/Makefile
> @@ -10,6 +10,7 @@ obj-$(CONFIG_TRANS_TABLE)	+= trans_pgd.o
>  obj-$(CONFIG_TRANS_TABLE)	+= trans_pgd-asm.o
>  obj-$(CONFIG_DEBUG_VIRTUAL)	+= physaddr.o
>  obj-$(CONFIG_ARM64_MTE)		+= mteswap.o
> +obj-$(CONFIG_ARM64_MTE_COMP)	+= mtecomp.o
>  KASAN_SANITIZE_physaddr.o	+= n
>  
>  obj-$(CONFIG_KASAN)		+= kasan_init.o
> diff --git a/arch/arm64/mm/mtecomp.c b/arch/arm64/mm/mtecomp.c
> new file mode 100644
> index 0000000000000..50a379c035aee
> --- /dev/null
> +++ b/arch/arm64/mm/mtecomp.c
> @@ -0,0 +1,406 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +/*
> + * MTE tag compression algorithm.
> + * Proposed by Evgenii Stepanov <eugenis@xxxxxxxxxx>
> + */
> +
> +/*
> + * EA0 stands for "Evgenii's Algorithm 0", as the initial proposal contained two
> + * compression algorithms.

This is the 4th time I see mr. Stepanov's credentials in the patch.
I've no doubts he's a worthy gentleman but please avoid mentioning
people in source code. Suggested-by is enough. IIRC, the rule for
that exists for about decade.

For the purpose of namespacing, the mte_compress/mte_decompress would
sound better.

> + *
> + * The algorithm attempts to compress a 128-byte (MTE_GRANULES_PER_PAGE / 2)
> + * array of tags into a smaller byte sequence that can be stored in a
> + * 16-, 32-, or 64-byte buffer. A special case is storing the tags inline in
> + * an 8-byte pointer.
> + *
> + * We encapsulate tag storage memory management in this module, because it is
> + * tightly coupled with the pointer representation.
> + *   ea0_compress(*tags) takes a 128-byte buffer and returns an opaque value
> + *     that can be stored in Xarray
> + *   ea0_decompress(*ptr, *tags) takes the opaque value and loads the tags into
> + *     the provided 128-byte buffer.
> + *
> + * The compression algorithm works as follows.
> + *
> + * 1. The input array of 128 bytes is transformed into tag ranges (two arrays:
> + *    @r_tags containing tag values and @r_sizes containing range lengths) by
> + *    ea0_tags_to_ranges(). Note that @r_sizes sums up to 256.
> + *
> + * 2. Depending on the number N of ranges, the following storage class is picked:
> + *            N <= 6:  8 bytes (inline case, no allocation required);
> + *       6 < N <= 11: 16 bytes
> + *      11 < N <= 23: 32 bytes
> + *      23 < N <= 46: 64 bytes
> + *      46 < N:       128 bytes (no compression will be performed)
> + *
> + * 3. The number of the largest element of @r_sizes is stored in @largest_idx.
> + *    The element itself is thrown away from @r_sizes, because it can be
> + *    reconstructed from the sum of the remaining elements. Note that now none
> + *    of the remaining @r_sizes elements is greater than 127.
> + *
> + * 4. For the inline case, the following values are stored in the 8-byte handle:
> + *       largest_idx : i4
> + *      r_tags[0..5] : i4 x 6
> + *     r_sizes[0..4] : i7 x 5
> + *    (if N is less than 6, @r_tags and @r_sizes are padded up with zero values)
> + *
> + *    Because @largest_idx is <= 5, bit 63 of the handle is always 0 (so it can
> + *    be stored in the Xarray), and bits 62..60 cannot all be 1, so it can be
> + *    distinguished from a kernel pointer.

I honestly tried to understand... For example, what the
        'r_sizes[0..4] : i7 x 5'
means? An array of 5 elements, 17 bits each? But it's alone greater
than size of pointer... Oh gosh...

Moreover, MTE tags are all 4-bits.

It seems like text without pictures is above my mental abilities. Can you
please illustrate it? For example, from the #4, I (hopefully correctly)
realized that:

Inline frame format:
   0                                                    60       62  63
 +---------------------------------------------------------------------+
 |              No idea what happens here             |    Lidx    | X |
 +---------------------------------------------------------------------+
 63    : X      :  RAZ : Reserved for Xarray.
 60-62 : Lidx   : 0..5 : Largest element index. 
                     6 : Reserved
                     7 : Invalid handler

> + *
> + * 5. For the out-of-line case, the storage is allocated from one of the
> + *    "mte-tags-{16,32,64,128}" kmem caches. The resulting pointer is aligned
> + *    on 8 bytes, so its bits 2..0 can be used to store the size class:
> + *     - 0 for 128 bytes
> + *     - 1 for 16
> + *     - 2 for 32
> + *     - 4 for 64.
> + *    Bit 63 of the pointer is zeroed out, so that it can be stored in Xarray.
> + *
> + * 6. The data layout in the allocated storage is as follows:
> + *         largest_idx : i6
> + *        r_tags[0..N] : i4 x N
> + *     r_sizes[0..N-1] : i7 x (N-1)
> + *
> + * The decompression algorithm performs the steps below.
> + *
> + * 1. Decide if data is stored inline (bits 62..60 of the handle != 0b111) or
> + *    out-of line.
> + *
> + * 2. For the inline case, treat the handle itself as the input buffer.
> + *
> + * 3. For the out-of-line case, look at bits 2..0 of the handle to understand
> + *    the input buffer length. To obtain the pointer to the input buffer, unset
> + *    bits 2..0 of the handle and set bit 63.
> + *
> + * 4. If the input buffer is 128 byte long, copy its contents to the output
> + *    buffer.
> + *
> + * 5. Otherwise, read @largest_idx, @r_tags and @r_sizes from the input buffer.
> + *    Calculate the removed largest element of @r_sizes:
> + *      largest = 256 - sum(r_sizes)
> + *    and insert it into @r_sizes at position @largest_idx.
> + *
> + * 6. While @r_sizes[i] > 0, add a 4-bit value @r_tags[i] to the output buffer
> + *    @r_sizes[i] times.
> + */
> +

Because of the size, I believe this comment is worth to put in Docs,
moreover we already have "Documentation/arch/arm64/memory-tagging-extension.rst"
Why don't you add an 'MTE Compression' section in there?

> +#include <linux/bits.h>
> +#include <linux/bitmap.h>
> +#include <linux/gfp.h>
> +#include <linux/module.h>
> +#include <linux/slab.h>
> +#include <linux/swab.h>
> +#include <linux/string.h>
> +#include <linux/types.h>

Nit: Alphabetic order?

Andy is right, bitmap.h includes bit.h, no need to include both. And if
your code will get broken one day, it's a bitmap maintainers' work to fix.

> +
> +#include <asm/mtecomp.h>
> +
> +/* The handle must fit into an Xarray value. */
> +#define HANDLE_MASK GENMASK_ULL(62, 0)
> +
> +/* Out-of-line handles have 0b111 in bits 62..60. */
> +#define NOINLINE_MASK GENMASK_ULL(62, 60)
> +
> +/* Cache index is stored in the lowest pointer bits. */
> +#define CACHE_ID_MASK GENMASK_ULL(2, 0)
> +
> +/*
> + * Four separate caches to store out-of-line data:
> + *  0: mte-tags-128
> + *  1: mte-tags-16
> + *  2: mte-tags-32
> + *  3: mte-tags-64
> + */
> +#define NUM_CACHES 4
> +static struct kmem_cache *mtecomp_caches[NUM_CACHES];
> +
> +/*
> + * Sizes of compressed values.
> + */
> +#define BITS_PER_TAG 4
> +#define BITS_PER_SIZE 7
> +#define BITS_PER_LARGEST_IDX_INLINE 4
> +#define BITS_PER_LARGEST_IDX 6

But in the comment you say that largest index in inline frame is 3-bits long.

> +
> +/* Translate allocation size into mtecomp_caches[] index. */
> +static int ea0_size_to_cache_id(int len)

Here and everywhere, do you need signed values? If not, unsigned int.

> +{
> +	if (len < 128)
> +		return fls(len >> 4);
> +	return 0;
> +}
> +
> +/* Translate mtecomp_caches[] index into allocation size. */
> +static int ea0_cache_id_to_size(int id)
> +{
> +	if (id == 0)
> +		return 128;
> +	return 8 << id;
> +}
> +
> +/* Transform tags into tag ranges. */
> +void ea0_tags_to_ranges(u8 *tags, u8 *out_tags, short *out_sizes, int *out_len)
> +{
> +	u8 prev_tag = U8_MAX;
> +	int cur_idx = -1;
> +	u8 cur_tag;
> +	int i, j;
> +
> +	memset(out_tags, 0, array_size(*out_len, sizeof(*out_tags)));
> +	memset(out_sizes, 0, array_size(*out_len, sizeof(*out_sizes)));
> +
> +	for (i = 0; i < MTE_PAGE_TAG_STORAGE; i++) {
> +		for (j = 0; j < 2; j++) {
> +			cur_tag = j ? (tags[i] % 16) : (tags[i] / 16);
> +			if (cur_tag == prev_tag) {
> +				out_sizes[cur_idx]++;
> +			} else {
> +				cur_idx++;
> +				prev_tag = cur_tag;
> +				out_tags[cur_idx] = prev_tag;
> +				out_sizes[cur_idx] = 1;
> +			}
> +		}
> +	}
> +	*out_len = cur_idx + 1;
> +}
> +EXPORT_SYMBOL_NS(ea0_tags_to_ranges, MTECOMP);
> +
> +/* Transform tag ranges back into tags. */
> +void ea0_ranges_to_tags(u8 *r_tags, short *r_sizes, int r_len, u8 *tags)
> +{
> +	int i, j, pos = 0;
> +	u8 prev;
> +
> +	for (i = 0; i < r_len; i++) {
> +		for (j = 0; j < r_sizes[i]; j++) {
> +			if (pos % 2)
> +				tags[pos / 2] = (prev << 4) | r_tags[i];
> +			else
> +				prev = r_tags[i];
> +			pos++;
> +		}
> +	}
> +}
> +EXPORT_SYMBOL_NS(ea0_ranges_to_tags, MTECOMP);

Because I didn't understand the compressed frame format, not sure I
can understand this logic...

> +
> +/* Translate @num_ranges into the allocation size needed to hold them. */
> +static int ea0_alloc_size(int num_ranges)
> +{
> +	if (num_ranges <= 6)
> +		return 8;
> +	if (num_ranges <= 11)
> +		return 16;
> +	if (num_ranges <= 23)
> +		return 32;
> +	if (num_ranges <= 46)
> +		return 64;
> +	return 128;
> +}
> +
> +/* Translate allocation size into maximum number of ranges that it can hold. */
> +static int ea0_size_to_ranges(int size)
> +{
> +	switch (size) {
> +	case 8:
> +		return 6;
> +	case 16:
> +		return 11;
> +	case 32:
> +		return 23;
> +	case 64:
> +		return 46;
> +	default:
> +		return 0;
> +	}
> +}

I wonder if there's a math formula here? Can you explain where from
those numbers come?

> +#define RANGES_INLINE ea0_size_to_ranges(8)

Maybe

 #define RANGES_INLINE (6)

> +
> +/* Is the data stored inline in the handle itself? */
> +static bool ea0_is_inline(unsigned long handle)
> +{
> +	return (handle & NOINLINE_MASK) != NOINLINE_MASK;
> +}
> +
> +/* Get the size of the buffer backing @handle. */
> +int ea0_storage_size(unsigned long handle)
> +{
> +	if (ea0_is_inline(handle))
> +		return 8;
> +	return ea0_cache_id_to_size(handle & CACHE_ID_MASK);
> +}
> +EXPORT_SYMBOL_NS(ea0_storage_size, MTECOMP);
> +
> +static void bitmap_write(unsigned long *bitmap, unsigned long value,
> +			 unsigned long *pos, unsigned long bits)

Please don't steal prefixes. But the idea is good. For the next
iteration, let's rename bitmap_set_value() to bitmap_write()?
So that your function will be an mte_bitmap_write().

Thanks,
Yury

> +{
> +	bitmap_set_value(bitmap, value, *pos, bits);
> +	*pos += bits;
> +}
> +
> +/* Compress ranges into the buffer that can accommodate up to max_ranges. */
> +static void ea0_compress_to_buf(int len, u8 *tags, short *sizes,
> +				unsigned long *bitmap, int max_ranges)
> +{
> +	unsigned long bit_pos = 0, l_bits;
> +	int largest_idx = -1, i;
> +	short largest = 0;
> +
> +	for (i = 0; i < len; i++) {
> +		if (sizes[i] > largest) {
> +			largest = sizes[i];
> +			largest_idx = i;
> +		}
> +	}
> +	l_bits = (max_ranges == RANGES_INLINE) ? BITS_PER_LARGEST_IDX_INLINE :
> +						 BITS_PER_LARGEST_IDX;
> +	bitmap_write(bitmap, largest_idx, &bit_pos, l_bits);
> +	for (i = 0; i < len; i++)
> +		bitmap_write(bitmap, tags[i], &bit_pos, BITS_PER_TAG);
> +	for (i = len; i < max_ranges; i++)
> +		bitmap_write(bitmap, 0, &bit_pos, BITS_PER_TAG);
> +	for (i = 0; i < len; i++) {
> +		if (i == largest_idx)
> +			continue;
> +		bitmap_write(bitmap, sizes[i], &bit_pos, BITS_PER_SIZE);
> +	}
> +	for (i = len; i < max_ranges; i++)
> +		bitmap_write(bitmap, 0, &bit_pos, BITS_PER_SIZE);
> +}
> +
> +/* Compress @tags and return a handle. */
> +unsigned long ea0_compress(u8 *tags)
> +{
> +	int alloc_size, cache_id;
> +	struct kmem_cache *cache;
> +	short r_sizes[256];
> +	u8 r_tags[256];
> +	int r_len = ARRAY_SIZE(r_tags);
> +	unsigned long *storage;
> +	/*
> +	 * ea0_compress_to_buf() only initializes the bits that ea0_decompress()
> +	 * will read. But when the tags are stored in the handle itself, it must
> +	 * have all its bits initialized.
> +	 */
> +	unsigned long result = 0;
> +
> +	ea0_tags_to_ranges(tags, r_tags, r_sizes, &r_len);
> +	alloc_size = ea0_alloc_size(r_len);
> +	if (alloc_size == 8) {
> +		ea0_compress_to_buf(r_len, r_tags, r_sizes, &result,
> +				    RANGES_INLINE);
> +		return result;
> +	}
> +	cache_id = ea0_size_to_cache_id(alloc_size);
> +	cache = mtecomp_caches[cache_id];
> +	storage = kmem_cache_alloc(cache, GFP_KERNEL);
> +	if (alloc_size < 128) {
> +		/* alloc_size is always a multiple of sizeof(unsigned long). */
> +		ea0_compress_to_buf(r_len, r_tags, r_sizes, storage,
> +				    ea0_size_to_ranges(alloc_size));
> +		return ((unsigned long)storage | cache_id) & HANDLE_MASK;
> +	}
> +	memcpy(storage, tags, alloc_size);
> +	return (unsigned long)storage & HANDLE_MASK;
> +}
> +EXPORT_SYMBOL_NS(ea0_compress, MTECOMP);
> +
> +static unsigned long bitmap_read(const unsigned long *bitmap,
> +				 unsigned long *pos, unsigned long bits)
> +{
> +	unsigned long result;
> +
> +	result = bitmap_get_value(bitmap, *pos, bits);
> +	*pos += bits;
> +	return result;
> +}
> +
> +/* Decompress the contents of the given buffer into @tags. */
> +static bool ea0_decompress_from_buf(const unsigned long *bitmap, int max_ranges,
> +				    u8 *tags)
> +{
> +	int largest_idx, i;
> +	short r_sizes[46], sum = 0;
> +	u8 r_tags[46];
> +	unsigned long bit_pos = 0, l_bits;
> +
> +	l_bits = (max_ranges == RANGES_INLINE) ? BITS_PER_LARGEST_IDX_INLINE :
> +						 BITS_PER_LARGEST_IDX;
> +	largest_idx = bitmap_read(bitmap, &bit_pos, l_bits);
> +	for (i = 0; i < max_ranges; i++)
> +		r_tags[i] = bitmap_read(bitmap, &bit_pos, BITS_PER_TAG);
> +	for (i = 0; i < max_ranges; i++) {
> +		if (i == largest_idx)
> +			continue;
> +		r_sizes[i] = bitmap_read(bitmap, &bit_pos, BITS_PER_SIZE);
> +		if (!r_sizes[i]) {
> +			max_ranges = i;
> +			break;
> +		}
> +		sum += r_sizes[i];
> +	}
> +	if (sum >= 256)
> +		return false;
> +	r_sizes[largest_idx] = 256 - sum;
> +	ea0_ranges_to_tags(r_tags, r_sizes, max_ranges, tags);
> +	return true;
> +}
> +
> +/* Get pointer to the out-of-line storage from a handle. */
> +static void *ea0_storage(unsigned long handle)
> +{
> +	if (ea0_is_inline(handle))
> +		return NULL;
> +	return (void *)((handle & (~CACHE_ID_MASK)) | BIT_ULL(63));
> +}
> +
> +/* Decompress tags from the buffer referenced by @handle. */
> +bool ea0_decompress(unsigned long handle, u8 *tags)
> +{
> +	unsigned long *storage = ea0_storage(handle);
> +	int size = ea0_storage_size(handle);
> +
> +	if (size == 128) {
> +		memcpy(tags, storage, size);
> +		return true;
> +	}
> +	if (size == 8)
> +		return ea0_decompress_from_buf(&handle, RANGES_INLINE, tags);
> +	return ea0_decompress_from_buf(storage, ea0_size_to_ranges(size), tags);
> +}
> +EXPORT_SYMBOL_NS(ea0_decompress, MTECOMP);
> +
> +/* Release the memory referenced by @handle. */
> +void ea0_release_handle(unsigned long handle)
> +{
> +	void *storage = ea0_storage(handle);
> +	int size = ea0_storage_size(handle);
> +	struct kmem_cache *c;
> +
> +	if (!storage)
> +		return;
> +
> +	c = mtecomp_caches[ea0_size_to_cache_id(size)];
> +	kmem_cache_free(c, storage);
> +}
> +EXPORT_SYMBOL_NS(ea0_release_handle, MTECOMP);
> +
> +/* Set up mtecomp_caches[]. */
> +static int mtecomp_init(void)
> +{
> +	char name[16];
> +	int size;
> +	int i;
> +
> +	BUILD_BUG_ON(MTE_PAGE_TAG_STORAGE != 128);
> +	for (i = 0; i < NUM_CACHES; i++) {
> +		size = ea0_cache_id_to_size(i);
> +		snprintf(name, ARRAY_SIZE(name), "mte-tags-%d", size);
> +		mtecomp_caches[i] =
> +			kmem_cache_create(name, size, size, 0, NULL);
> +	}
> +	return 0;
> +}
> +module_init(mtecomp_init);
> -- 
> 2.41.0.255.g8b1d071c50-goog