[PATCH v10-mte 0/7] Implement MTE tag compression for swapped pages

[PATCH v10-mte 0/7] Implement MTE tag compression for swapped pages

[Alexander Potapenko]

Currently, when MTE pages are swapped out, the tags are kept in the
memory, occupying PAGE_SIZE/32 bytes per page. This is especially
problematic for devices that use zram-backed in-memory swap, because
tags stored uncompressed in the heap effectively reduce the available
amount of swap memory.

The RLE-based algorithm suggested by Evgenii Stepanov and implemented in
this patch series is able to efficiently compress fixed-size tag buffers,
resulting in practical compression ratio of 2x. In many cases it is
possible to store the compressed data in 63-bit Xarray values, resulting
in no extra memory allocations.

To implement compression/decompression, we also extend <linux/bitmap.h>
with methods to read/write bit values at arbitrary places in the map.

This patch series depends on "lib/bitmap: add bitmap_{read,write}()"
(https://lore.kernel.org/linux-arm-kernel/20231030153210.139512-1-glider@google.com/T/)
that is mailed separately.

v10-mte:
 - add the bitmap_read()/bitmap_write() patches
   (https://lore.kernel.org/lkml/20231109151106.2385155-1-glider@google.com/)
   back to this series
 - add Acked-by: tags to MTE patches

v9:
 - split off the stats collection code into a separate patch in the
   series (as suggested by Yury Norov)

v8:
 - split off the bitmap_read()/bitmap_write() series
 - simplified the compression logic (only compress data if it fits into
   a pointer)

v7:
 - fixed comments by Yury Norov, Andy Shevchenko, Rasmus Villemoes
 - added perf tests for bitmap_read()/bitmap_write()
 - more efficient bitmap_write() implementation (meant to be sent in v5)

v6:
 - fixed comments by Yury Norov
 - fixed handling of sizes divisible by MTE_GRANULES_PER_PAGE / 2
   (caught while testing on a real device)

v5:
 - fixed comments by Andy Shevchenko, Catalin Marinas, and Yury Norov
 - added support for 16K- and 64K pages
 - more efficient bitmap_write() implementation

v4:
 - fixed a bunch of comments by Andy Shevchenko and Yury Norov
 - added Documentation/arch/arm64/mte-tag-compression.rst

v3:
 - as suggested by Andy Shevchenko, use
   bitmap_get_value()/bitmap_set_value() written by Syed Nayyar Waris
 - switched to unsigned long to reduce typecasts
 - simplified the compression code

v2:
 - as suggested by Yuri Norov, replace the poorly implemented struct
   bitq with <linux/bitmap.h>


Alexander Potapenko (6):
  lib/test_bitmap: add tests for bitmap_{read,write}()
  lib/test_bitmap: use pr_info() for non-error messages
  arm64: mte: implement CONFIG_ARM64_MTE_COMP
  arm64: mte: add a test for MTE tags compression
  arm64: mte: add compression support to mteswap.c
  arm64: mte: implement CONFIG_ARM64_MTE_SWAP_STATS

Syed Nayyar Waris (1):
  lib/bitmap: add bitmap_{read,write}()

 Documentation/arch/arm64/index.rst            |   1 +
 .../arch/arm64/mte-tag-compression.rst        | 166 ++++++++
 arch/arm64/Kconfig                            |  37 ++
 arch/arm64/include/asm/mtecomp.h              |  39 ++
 arch/arm64/mm/Makefile                        |   2 +
 arch/arm64/mm/mtecomp.c                       | 257 +++++++++++++
 arch/arm64/mm/mtecomp.h                       |  12 +
 arch/arm64/mm/mteswap.c                       | 110 +++++-
 arch/arm64/mm/test_mtecomp.c                  | 364 ++++++++++++++++++
 include/linux/bitmap.h                        |  77 ++++
 lib/test_bitmap.c                             | 185 ++++++++-
 11 files changed, 1237 insertions(+), 13 deletions(-)
 create mode 100644 Documentation/arch/arm64/mte-tag-compression.rst
 create mode 100644 arch/arm64/include/asm/mtecomp.h
 create mode 100644 arch/arm64/mm/mtecomp.c
 create mode 100644 arch/arm64/mm/mtecomp.h
 create mode 100644 arch/arm64/mm/test_mtecomp.c

-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 1/7] lib/bitmap: add bitmap_{read,write}()

[Alexander Potapenko]

From: Syed Nayyar Waris <syednwaris@gmail.com>

The two new functions allow reading/writing values of length up to
BITS_PER_LONG bits at arbitrary position in the bitmap.

The code was taken from "bitops: Introduce the for_each_set_clump macro"
by Syed Nayyar Waris with a number of changes and simplifications:
 - instead of using roundup(), which adds an unnecessary dependency
   on <linux/math.h>, we calculate space as BITS_PER_LONG-offset;
 - indentation is reduced by not using else-clauses (suggested by
   checkpatch for bitmap_get_value());
 - bitmap_get_value()/bitmap_set_value() are renamed to bitmap_read()
   and bitmap_write();
 - some redundant computations are omitted.

Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Syed Nayyar Waris <syednwaris@gmail.com>
Signed-off-by: William Breathitt Gray <william.gray@linaro.org>
Link: https://lore.kernel.org/lkml/fe12eedf3666f4af5138de0e70b67a07c7f40338.1592224129.git.syednwaris@gmail.com/
Suggested-by: Yury Norov <yury.norov@gmail.com>
Co-developed-by: Alexander Potapenko <glider@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

---

v10-mte:
 - send this patch together with the "Implement MTE tag compression for
   swapped pages"

Revisions v8-v12 of bitmap patches were reviewed separately from the
"Implement MTE tag compression for swapped pages" series
(https://lore.kernel.org/lkml/20231109151106.2385155-1-glider@google.com/)

This patch was previously called "lib/bitmap: add
bitmap_{set,get}_value()"
(https://lore.kernel.org/lkml/20230720173956.3674987-2-glider@google.com/)

v11:
 - rearrange whitespace as requested by Andy Shevchenko,
   add Reviewed-by:, update a comment

v10:
 - update comments as requested by Andy Shevchenko

v8:
 - as suggested by Andy Shevchenko, handle reads/writes of more than
   BITS_PER_LONG bits, add a note for 32-bit systems

v7:
 - Address comments by Yury Norov, Andy Shevchenko, Rasmus Villemoes:
   - update code comments;
   - get rid of GENMASK();
   - s/assign_bit/__assign_bit;
   - more vertical whitespace for better readability;
 - more compact code for bitmap_write() (now for real)

v6:
 - As suggested by Yury Norov, do not require bitmap_read(..., 0) to
   return 0.

v5:
 - Address comments by Yury Norov:
   - updated code comments and patch title/description
   - replace GENMASK(nbits - 1, 0) with BITMAP_LAST_WORD_MASK(nbits)
   - more compact bitmap_write() implementation

v4:
 - Address comments by Andy Shevchenko and Yury Norov:
   - prevent passing values >= 64 to GENMASK()
   - fix commit authorship
   - change comments
   - check for unlikely(nbits==0)
   - drop unnecessary const declarations
   - fix kernel-doc comments
   - rename bitmap_{get,set}_value() to bitmap_{read,write}()
---
 include/linux/bitmap.h | 77 ++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 77 insertions(+)

diff --git a/include/linux/bitmap.h b/include/linux/bitmap.h
index 99451431e4d65..7ca0379be8c13 100644
--- a/include/linux/bitmap.h
+++ b/include/linux/bitmap.h
@@ -79,6 +79,10 @@ struct device;
  *  bitmap_to_arr64(buf, src, nbits)            Copy nbits from buf to u64[] dst
  *  bitmap_get_value8(map, start)               Get 8bit value from map at start
  *  bitmap_set_value8(map, value, start)        Set 8bit value to map at start
+ *  bitmap_read(map, start, nbits)              Read an nbits-sized value from
+ *                                              map at start
+ *  bitmap_write(map, value, start, nbits)      Write an nbits-sized value to
+ *                                              map at start
  *
  * Note, bitmap_zero() and bitmap_fill() operate over the region of
  * unsigned longs, that is, bits behind bitmap till the unsigned long
@@ -636,6 +640,79 @@ static inline void bitmap_set_value8(unsigned long *map, unsigned long value,
 	map[index] |= value << offset;
 }
 
+/**
+ * bitmap_read - read a value of n-bits from the memory region
+ * @map: address to the bitmap memory region
+ * @start: bit offset of the n-bit value
+ * @nbits: size of value in bits, nonzero, up to BITS_PER_LONG
+ *
+ * Returns: value of @nbits bits located at the @start bit offset within the
+ * @map memory region. For @nbits = 0 and @nbits > BITS_PER_LONG the return
+ * value is undefined.
+ */
+static inline unsigned long bitmap_read(const unsigned long *map,
+					unsigned long start,
+					unsigned long nbits)
+{
+	size_t index = BIT_WORD(start);
+	unsigned long offset = start % BITS_PER_LONG;
+	unsigned long space = BITS_PER_LONG - offset;
+	unsigned long value_low, value_high;
+
+	if (unlikely(!nbits || nbits > BITS_PER_LONG))
+		return 0;
+
+	if (space >= nbits)
+		return (map[index] >> offset) & BITMAP_LAST_WORD_MASK(nbits);
+
+	value_low = map[index] & BITMAP_FIRST_WORD_MASK(start);
+	value_high = map[index + 1] & BITMAP_LAST_WORD_MASK(start + nbits);
+	return (value_low >> offset) | (value_high << space);
+}
+
+/**
+ * bitmap_write - write n-bit value within a memory region
+ * @map: address to the bitmap memory region
+ * @value: value to write, clamped to nbits
+ * @start: bit offset of the n-bit value
+ * @nbits: size of value in bits, nonzero, up to BITS_PER_LONG.
+ *
+ * bitmap_write() behaves as-if implemented as @nbits calls of __assign_bit(),
+ * i.e. bits beyond @nbits are ignored:
+ *
+ *   for (bit = 0; bit < nbits; bit++)
+ *           __assign_bit(start + bit, bitmap, val & BIT(bit));
+ *
+ * For @nbits == 0 and @nbits > BITS_PER_LONG no writes are performed.
+ */
+static inline void bitmap_write(unsigned long *map, unsigned long value,
+				unsigned long start, unsigned long nbits)
+{
+	size_t index;
+	unsigned long offset;
+	unsigned long space;
+	unsigned long mask;
+	bool fit;
+
+	if (unlikely(!nbits || nbits > BITS_PER_LONG))
+		return;
+
+	mask = BITMAP_LAST_WORD_MASK(nbits);
+	value &= mask;
+	offset = start % BITS_PER_LONG;
+	space = BITS_PER_LONG - offset;
+	fit = space >= nbits;
+	index = BIT_WORD(start);
+
+	map[index] &= (fit ? (~(mask << offset)) : ~BITMAP_FIRST_WORD_MASK(start));
+	map[index] |= value << offset;
+	if (fit)
+		return;
+
+	map[index + 1] &= BITMAP_FIRST_WORD_MASK(start + nbits);
+	map[index + 1] |= (value >> space);
+}
+
 #endif /* __ASSEMBLY__ */
 
 #endif /* __LINUX_BITMAP_H */
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 1/7] lib/bitmap: add bitmap_{read,write}()

[Yury Norov]

On Thu, Dec 14, 2023 at 12:06:33PM +0100, Alexander Potapenko wrote:
> From: Syed Nayyar Waris <syednwaris@gmail.com>
> 
> The two new functions allow reading/writing values of length up to
> BITS_PER_LONG bits at arbitrary position in the bitmap.
> 
> The code was taken from "bitops: Introduce the for_each_set_clump macro"
> by Syed Nayyar Waris with a number of changes and simplifications:
>  - instead of using roundup(), which adds an unnecessary dependency
>    on <linux/math.h>, we calculate space as BITS_PER_LONG-offset;
>  - indentation is reduced by not using else-clauses (suggested by
>    checkpatch for bitmap_get_value());
>  - bitmap_get_value()/bitmap_set_value() are renamed to bitmap_read()
>    and bitmap_write();
>  - some redundant computations are omitted.
> 
> Cc: Arnd Bergmann <arnd@arndb.de>
> Signed-off-by: Syed Nayyar Waris <syednwaris@gmail.com>
> Signed-off-by: William Breathitt Gray <william.gray@linaro.org>
> Link: https://lore.kernel.org/lkml/fe12eedf3666f4af5138de0e70b67a07c7f40338.1592224129.git.syednwaris@gmail.com/
> Suggested-by: Yury Norov <yury.norov@gmail.com>
> Co-developed-by: Alexander Potapenko <glider@google.com>
> Signed-off-by: Alexander Potapenko <glider@google.com>
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

Acked-by: Yury Norov <yury.norov@gmail.com>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 2/7] lib/test_bitmap: add tests for bitmap_{read,write}()

[Alexander Potapenko]

Add basic tests ensuring that values can be added at arbitrary positions
of the bitmap, including those spanning into the adjacent unsigned
longs.

Two new performance tests, test_bitmap_read_perf() and
test_bitmap_write_perf(), can be used to assess future performance
improvements of bitmap_read() and bitmap_write():

[    0.431119][    T1] test_bitmap: Time spent in test_bitmap_read_perf:	615253
[    0.433197][    T1] test_bitmap: Time spent in test_bitmap_write_perf:	916313

(numbers from a Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz machine running
QEMU).

Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

---
v10-mte:
 - send this patch together with the "Implement MTE tag compression for
   swapped pages"

Revisions v8-v12 of bitmap patches were reviewed separately from the
"Implement MTE tag compression for swapped pages" series
(https://lore.kernel.org/lkml/20231109151106.2385155-1-glider@google.com/)

This patch was previously called
"lib/test_bitmap: add tests for bitmap_{set,get}_value()"
(https://lore.kernel.org/lkml/20230720173956.3674987-3-glider@google.com/)
and
"lib/test_bitmap: add tests for bitmap_{set,get}_value_unaligned"
(https://lore.kernel.org/lkml/20230713125706.2884502-3-glider@google.com/)

v12:
 - as suggested by Alexander Lobakin, replace expect_eq_uint() with
   expect_eq_ulong() and a cast

v9:
 - use WRITE_ONCE() to prevent optimizations in test_bitmap_read_perf()
 - update patch description

v8:
 - as requested by Andy Shevchenko, add tests for reading/writing
   sizes > BITS_PER_LONG

v7:
 - as requested by Yury Norov, add performance tests for bitmap_read()
   and bitmap_write()

v6:
 - use bitmap API to initialize test bitmaps
 - as requested by Yury Norov, do not check the return value of
   bitmap_read(..., 0)
 - fix a compiler warning on 32-bit systems

v5:
 - update patch title
 - address Yury Norov's comments:
   - rename the test cases
   - factor out test_bitmap_write_helper() to test writing over
     different background patterns;
   - add a test case copying a nontrivial value bit-by-bit;
   - drop volatile

v4:
 - Address comments by Andy Shevchenko: added Reviewed-by: and a link to
   the previous discussion
 - Address comments by Yury Norov:
   - expand the bitmap to catch more corner cases
   - add code testing that bitmap_set_value() does not touch adjacent
     bits
   - add code testing the nbits==0 case
   - rename bitmap_{get,set}_value() to bitmap_{read,write}()

v3:
 - switch to using bitmap_{set,get}_value()
 - change the expected bit pattern in test_set_get_value(),
   as the test was incorrectly assuming 0 is the LSB.
---
 lib/test_bitmap.c | 179 ++++++++++++++++++++++++++++++++++++++++++++--
 1 file changed, 172 insertions(+), 7 deletions(-)

diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c
index 65f22c2578b06..46c0154680772 100644
--- a/lib/test_bitmap.c
+++ b/lib/test_bitmap.c
@@ -60,18 +60,17 @@ static const unsigned long exp3_1_0[] __initconst = {
 };
 
 static bool __init
-__check_eq_uint(const char *srcfile, unsigned int line,
-		const unsigned int exp_uint, unsigned int x)
+__check_eq_ulong(const char *srcfile, unsigned int line,
+		 const unsigned long exp_ulong, unsigned long x)
 {
-	if (exp_uint != x) {
-		pr_err("[%s:%u] expected %u, got %u\n",
-			srcfile, line, exp_uint, x);
+	if (exp_ulong != x) {
+		pr_err("[%s:%u] expected %lu, got %lu\n",
+			srcfile, line, exp_ulong, x);
 		return false;
 	}
 	return true;
 }
 
-
 static bool __init
 __check_eq_bitmap(const char *srcfile, unsigned int line,
 		  const unsigned long *exp_bmap, const unsigned long *bmap,
@@ -185,7 +184,8 @@ __check_eq_str(const char *srcfile, unsigned int line,
 		result;							\
 	})
 
-#define expect_eq_uint(...)		__expect_eq(uint, ##__VA_ARGS__)
+#define expect_eq_ulong(...)		__expect_eq(ulong, ##__VA_ARGS__)
+#define expect_eq_uint(x, y)		expect_eq_ulong((unsigned int)(x), (unsigned int)(y))
 #define expect_eq_bitmap(...)		__expect_eq(bitmap, ##__VA_ARGS__)
 #define expect_eq_pbl(...)		__expect_eq(pbl, ##__VA_ARGS__)
 #define expect_eq_u32_array(...)	__expect_eq(u32_array, ##__VA_ARGS__)
@@ -1245,6 +1245,168 @@ static void __init test_bitmap_const_eval(void)
 	BUILD_BUG_ON(~var != ~BIT(25));
 }
 
+/*
+ * Test bitmap should be big enough to include the cases when start is not in
+ * the first word, and start+nbits lands in the following word.
+ */
+#define TEST_BIT_LEN (1000)
+
+/*
+ * Helper function to test bitmap_write() overwriting the chosen byte pattern.
+ */
+static void __init test_bitmap_write_helper(const char *pattern)
+{
+	DECLARE_BITMAP(bitmap, TEST_BIT_LEN);
+	DECLARE_BITMAP(exp_bitmap, TEST_BIT_LEN);
+	DECLARE_BITMAP(pat_bitmap, TEST_BIT_LEN);
+	unsigned long w, r, bit;
+	int i, n, nbits;
+
+	/*
+	 * Only parse the pattern once and store the result in the intermediate
+	 * bitmap.
+	 */
+	bitmap_parselist(pattern, pat_bitmap, TEST_BIT_LEN);
+
+	/*
+	 * Check that writing a single bit does not accidentally touch the
+	 * adjacent bits.
+	 */
+	for (i = 0; i < TEST_BIT_LEN; i++) {
+		bitmap_copy(bitmap, pat_bitmap, TEST_BIT_LEN);
+		bitmap_copy(exp_bitmap, pat_bitmap, TEST_BIT_LEN);
+		for (bit = 0; bit <= 1; bit++) {
+			bitmap_write(bitmap, bit, i, 1);
+			__assign_bit(i, exp_bitmap, bit);
+			expect_eq_bitmap(exp_bitmap, bitmap,
+					 TEST_BIT_LEN);
+		}
+	}
+
+	/* Ensure writing 0 bits does not change anything. */
+	bitmap_copy(bitmap, pat_bitmap, TEST_BIT_LEN);
+	bitmap_copy(exp_bitmap, pat_bitmap, TEST_BIT_LEN);
+	for (i = 0; i < TEST_BIT_LEN; i++) {
+		bitmap_write(bitmap, ~0UL, i, 0);
+		expect_eq_bitmap(exp_bitmap, bitmap, TEST_BIT_LEN);
+	}
+
+	for (nbits = BITS_PER_LONG; nbits >= 1; nbits--) {
+		w = IS_ENABLED(CONFIG_64BIT) ? 0xdeadbeefdeadbeefUL
+					     : 0xdeadbeefUL;
+		w >>= (BITS_PER_LONG - nbits);
+		for (i = 0; i <= TEST_BIT_LEN - nbits; i++) {
+			bitmap_copy(bitmap, pat_bitmap, TEST_BIT_LEN);
+			bitmap_copy(exp_bitmap, pat_bitmap, TEST_BIT_LEN);
+			for (n = 0; n < nbits; n++)
+				__assign_bit(i + n, exp_bitmap, w & BIT(n));
+			bitmap_write(bitmap, w, i, nbits);
+			expect_eq_bitmap(exp_bitmap, bitmap, TEST_BIT_LEN);
+			r = bitmap_read(bitmap, i, nbits);
+			expect_eq_ulong(r, w);
+		}
+	}
+}
+
+static void __init test_bitmap_read_write(void)
+{
+	unsigned char *pattern[3] = {"", "all:1/2", "all"};
+	DECLARE_BITMAP(bitmap, TEST_BIT_LEN);
+	unsigned long zero_bits = 0, bits_per_long = BITS_PER_LONG;
+	unsigned long val;
+	int i, pi;
+
+	/*
+	 * Reading/writing zero bits should not crash the kernel.
+	 * READ_ONCE() prevents constant folding.
+	 */
+	bitmap_write(NULL, 0, 0, READ_ONCE(zero_bits));
+	/* Return value of bitmap_read() is undefined here. */
+	bitmap_read(NULL, 0, READ_ONCE(zero_bits));
+
+	/*
+	 * Reading/writing more than BITS_PER_LONG bits should not crash the
+	 * kernel. READ_ONCE() prevents constant folding.
+	 */
+	bitmap_write(NULL, 0, 0, READ_ONCE(bits_per_long) + 1);
+	/* Return value of bitmap_read() is undefined here. */
+	bitmap_read(NULL, 0, READ_ONCE(bits_per_long) + 1);
+
+	/*
+	 * Ensure that bitmap_read() reads the same value that was previously
+	 * written, and two consequent values are correctly merged.
+	 * The resulting bit pattern is asymmetric to rule out possible issues
+	 * with bit numeration order.
+	 */
+	for (i = 0; i < TEST_BIT_LEN - 7; i++) {
+		bitmap_zero(bitmap, TEST_BIT_LEN);
+
+		bitmap_write(bitmap, 0b10101UL, i, 5);
+		val = bitmap_read(bitmap, i, 5);
+		expect_eq_ulong(0b10101UL, val);
+
+		bitmap_write(bitmap, 0b101UL, i + 5, 3);
+		val = bitmap_read(bitmap, i + 5, 3);
+		expect_eq_ulong(0b101UL, val);
+
+		val = bitmap_read(bitmap, i, 8);
+		expect_eq_ulong(0b10110101UL, val);
+	}
+
+	for (pi = 0; pi < ARRAY_SIZE(pattern); pi++)
+		test_bitmap_write_helper(pattern[pi]);
+}
+
+static void __init test_bitmap_read_perf(void)
+{
+	DECLARE_BITMAP(bitmap, TEST_BIT_LEN);
+	unsigned int cnt, nbits, i;
+	unsigned long val;
+	ktime_t time;
+
+	bitmap_fill(bitmap, TEST_BIT_LEN);
+	time = ktime_get();
+	for (cnt = 0; cnt < 5; cnt++) {
+		for (nbits = 1; nbits <= BITS_PER_LONG; nbits++) {
+			for (i = 0; i < TEST_BIT_LEN; i++) {
+				if (i + nbits > TEST_BIT_LEN)
+					break;
+				/*
+				 * Prevent the compiler from optimizing away the
+				 * bitmap_read() by using its value.
+				 */
+				WRITE_ONCE(val, bitmap_read(bitmap, i, nbits));
+			}
+		}
+	}
+	time = ktime_get() - time;
+	pr_err("Time spent in %s:\t%llu\n", __func__, time);
+}
+
+static void __init test_bitmap_write_perf(void)
+{
+	DECLARE_BITMAP(bitmap, TEST_BIT_LEN);
+	unsigned int cnt, nbits, i;
+	unsigned long val = 0xfeedface;
+	ktime_t time;
+
+	bitmap_zero(bitmap, TEST_BIT_LEN);
+	time = ktime_get();
+	for (cnt = 0; cnt < 5; cnt++) {
+		for (nbits = 1; nbits <= BITS_PER_LONG; nbits++) {
+			for (i = 0; i < TEST_BIT_LEN; i++) {
+				if (i + nbits > TEST_BIT_LEN)
+					break;
+				bitmap_write(bitmap, val, i, nbits);
+			}
+		}
+	}
+	time = ktime_get() - time;
+	pr_err("Time spent in %s:\t%llu\n", __func__, time);
+}
+
+#undef TEST_BIT_LEN
+
 static void __init selftest(void)
 {
 	test_zero_clear();
@@ -1261,6 +1423,9 @@ static void __init selftest(void)
 	test_bitmap_cut();
 	test_bitmap_print_buf();
 	test_bitmap_const_eval();
+	test_bitmap_read_write();
+	test_bitmap_read_perf();
+	test_bitmap_write_perf();
 
 	test_find_nth_bit();
 	test_for_each_set_bit();
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 2/7] lib/test_bitmap: add tests for bitmap_{read,write}()

[Yury Norov]

On Thu, Dec 14, 2023 at 12:06:34PM +0100, Alexander Potapenko wrote:
> Add basic tests ensuring that values can be added at arbitrary positions
> of the bitmap, including those spanning into the adjacent unsigned
> longs.
> 
> Two new performance tests, test_bitmap_read_perf() and
> test_bitmap_write_perf(), can be used to assess future performance
> improvements of bitmap_read() and bitmap_write():
> 
> [    0.431119][    T1] test_bitmap: Time spent in test_bitmap_read_perf:	615253
> [    0.433197][    T1] test_bitmap: Time spent in test_bitmap_write_perf:	916313
> 
> (numbers from a Intel(R) Xeon(R) Gold 6154 CPU @ 3.00GHz machine running
> QEMU).
> 
> Signed-off-by: Alexander Potapenko <glider@google.com>
> Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
 
Acked-by: Yury Norov <yury.norov@gmail.com>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 3/7] lib/test_bitmap: use pr_info() for non-error messages

[Alexander Potapenko]

pr_err() messages may be treated as errors by some log readers, so let
us only use them for test failures. For non-error messages, replace them
with pr_info().

Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
Signed-off-by: Alexander Potapenko <glider@google.com>

---
v10-mte:
 - send this patch together with the "Implement MTE tag compression for
   swapped pages"
---
 lib/test_bitmap.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/lib/test_bitmap.c b/lib/test_bitmap.c
index 46c0154680772..a6e92cf5266af 100644
--- a/lib/test_bitmap.c
+++ b/lib/test_bitmap.c
@@ -507,7 +507,7 @@ static void __init test_bitmap_parselist(void)
 		}
 
 		if (ptest.flags & PARSE_TIME)
-			pr_err("parselist: %d: input is '%s' OK, Time: %llu\n",
+			pr_info("parselist: %d: input is '%s' OK, Time: %llu\n",
 					i, ptest.in, time);
 
 #undef ptest
@@ -546,7 +546,7 @@ static void __init test_bitmap_printlist(void)
 		goto out;
 	}
 
-	pr_err("bitmap_print_to_pagebuf: input is '%s', Time: %llu\n", buf, time);
+	pr_info("bitmap_print_to_pagebuf: input is '%s', Time: %llu\n", buf, time);
 out:
 	kfree(buf);
 	kfree(bmap);
@@ -624,7 +624,7 @@ static void __init test_bitmap_parse(void)
 		}
 
 		if (test.flags & PARSE_TIME)
-			pr_err("parse: %d: input is '%s' OK, Time: %llu\n",
+			pr_info("parse: %d: input is '%s' OK, Time: %llu\n",
 					i, test.in, time);
 	}
 }
@@ -1380,7 +1380,7 @@ static void __init test_bitmap_read_perf(void)
 		}
 	}
 	time = ktime_get() - time;
-	pr_err("Time spent in %s:\t%llu\n", __func__, time);
+	pr_info("Time spent in %s:\t%llu\n", __func__, time);
 }
 
 static void __init test_bitmap_write_perf(void)
@@ -1402,7 +1402,7 @@ static void __init test_bitmap_write_perf(void)
 		}
 	}
 	time = ktime_get() - time;
-	pr_err("Time spent in %s:\t%llu\n", __func__, time);
+	pr_info("Time spent in %s:\t%llu\n", __func__, time);
 }
 
 #undef TEST_BIT_LEN
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 3/7] lib/test_bitmap: use pr_info() for non-error messages

[Yury Norov]

On Thu, Dec 14, 2023 at 12:06:35PM +0100, Alexander Potapenko wrote:
> pr_err() messages may be treated as errors by some log readers, so let
> us only use them for test failures. For non-error messages, replace them
> with pr_info().
> 
> Suggested-by: Alexander Lobakin <aleksander.lobakin@intel.com>
> Signed-off-by: Alexander Potapenko <glider@google.com>

Acked-by: Yury Norov <yury.norov@gmail.com>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Alexander Potapenko]

The config implements the algorithm compressing memory tags for ARM MTE
during swapping.

The algorithm is based on RLE and specifically targets buffers of tags
corresponding to a single page. In many cases a buffer can be compressed
into 63 bits, making it possible to store it without additional memory
allocation.

Suggested-by: Evgenii Stepanov <eugenis@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

---
 v10-mte:
  - added Catalin's Acked-by:

 v8:
  - As suggested by Catalin Marinas, only compress tags if they can be
    stored inline. This simplifies the code drastically.
  - Update the documentation.
  - Split off patches introducing bitmap_read()/bitmap_write().

 v6:
  - shuffle bits in inline handles so that they can't be confused with
    canonical pointers;
  - use kmem_cache_zalloc() to allocate compressed storage
  - correctly handle range size overflow
  - minor documentation fixes, clarify the special cases

 v5:
  - make code PAGE_SIZE-agnostic, remove hardcoded constants, updated
    the docs
  - implement debugfs interface
  - Address comments by Andy Shevchenko:
    - update description of mtecomp.c
    - remove redundant assignments, simplify mte_tags_to_ranges()
    - various code simplifications
    - introduce mtecomp.h
    - add :export: to Documentation/arch/arm64/mte-tag-compression.rst

 v4:
  - Addressed comments by Andy Shevchenko:
    - expanded "MTE" to "Memory Tagging Extension" in Kconfig
    - fixed kernel-doc comments, moved them to C source
    - changed variables to unsigned where applicable
    - some code simplifications, fewer unnecessary assignments
    - added the mte_largest_idx_bits() helper
    - added namespace prefixes to all functions
    - added missing headers (but removed bits.h)
  - Addressed comments by Yury Norov:
    - removed test-only functions from mtecomp.h
    - dropped the algoritm name (all functions are now prefixed with
      "mte")
    - added more comments
    - got rid of MTE_RANGES_INLINE
    - renamed bitmap_{get,set}_value() to bitmap_{read,write}()
    - moved the big comment explaining the algorithm to
      Documentation/arch/arm64/mte-tag-compression.rst, expanded it,
      add a link to it from Documentation/arch/arm64/index.rst
    - removed hardcoded ranges from mte_alloc_size()/mte_size_to_ranges()

 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 Yury Norov, switched from struct bitq (which is
    not needed anymore) to <linux/bitmap.h>
  - add missing symbol exports
---
 Documentation/arch/arm64/index.rst            |   1 +
 .../arch/arm64/mte-tag-compression.rst        | 154 +++++++++++
 arch/arm64/Kconfig                            |  11 +
 arch/arm64/include/asm/mtecomp.h              |  39 +++
 arch/arm64/mm/Makefile                        |   1 +
 arch/arm64/mm/mtecomp.c                       | 257 ++++++++++++++++++
 arch/arm64/mm/mtecomp.h                       |  12 +
 7 files changed, 475 insertions(+)
 create mode 100644 Documentation/arch/arm64/mte-tag-compression.rst
 create mode 100644 arch/arm64/include/asm/mtecomp.h
 create mode 100644 arch/arm64/mm/mtecomp.c
 create mode 100644 arch/arm64/mm/mtecomp.h

diff --git a/Documentation/arch/arm64/index.rst b/Documentation/arch/arm64/index.rst
index d08e924204bf1..bf6c1583233a9 100644
--- a/Documentation/arch/arm64/index.rst
+++ b/Documentation/arch/arm64/index.rst
@@ -19,6 +19,7 @@ ARM64 Architecture
     legacy_instructions
     memory
     memory-tagging-extension
+    mte-tag-compression
     perf
     pointer-authentication
     ptdump
diff --git a/Documentation/arch/arm64/mte-tag-compression.rst b/Documentation/arch/arm64/mte-tag-compression.rst
new file mode 100644
index 0000000000000..8fe6b51a9db6d
--- /dev/null
+++ b/Documentation/arch/arm64/mte-tag-compression.rst
@@ -0,0 +1,154 @@
+.. SPDX-License-Identifier: GPL-2.0
+
+==================================================
+Tag Compression for Memory Tagging Extension (MTE)
+==================================================
+
+This document describes the algorithm used to compress memory tags used by the
+ARM Memory Tagging Extension (MTE).
+
+Introduction
+============
+
+MTE assigns tags to memory pages: for 4K pages those tags occupy 128 bytes
+(256 4-bit tags each corresponding to a 16-byte MTE granule), for 16K pages -
+512 bytes, for 64K pages - 2048 bytes. By default, MTE carves out 3.125% (1/16)
+of the available physical memory to store the tags.
+
+When MTE pages are saved to swap, their tags need to be stored in the kernel
+memory. If the system swap is used heavily, these tags may take a substantial
+portion of the physical memory. To reduce memory waste, ``CONFIG_ARM64_MTE_COMP``
+allows the kernel to store the tags in compressed form.
+
+Implementation details
+======================
+
+The algorithm attempts to compress an array of ``MTE_PAGE_TAG_STORAGE``
+tag bytes into a byte sequence that can be stored in an 8-byte pointer. If that
+is not possible, the data is stored uncompressed.
+
+Tag manipulation and storage
+----------------------------
+
+Tags for swapped pages are stored in an XArray that maps swap entries to 63-bit
+values (see ``arch/arm64/mm/mteswap.c``). Bit 0 of these values indicates how
+their contents should be treated:
+
+ - 0: value is a pointer to an uncompressed buffer allocated with kmalloc()
+   (always the case if ``CONFIG_ARM64_MTE_COMP=n``) with the highest bit set
+   to 0;
+ - 1: value contains compressed data.
+
+``arch/arm64/include/asm/mtecomp.h`` declares the following functions that
+manipulate with tags:
+
+- mte_compress() - compresses the given ``MTE_PAGE_TAG_STORAGE``-byte ``tags``
+  buffer into a pointer;
+- mte_decompress() - decompresses the tags from a pointer;
+- mte_is_compressed() - returns ``true`` iff the pointer passed to it should be
+  treated as compressed data.
+
+Tag compression
+---------------
+
+The compression algorithm is a variation of RLE (run-length encoding) and works
+as follows (we will be considering 4K pages and 128-byte tag buffers, but the
+same approach scales to 16K and 64K pages):
+
+1. The input array of 128 (``MTE_PAGE_TAG_STORAGE``) bytes is transformed into
+   tag ranges (two arrays: ``r_tags[]`` containing tag values and ``r_sizes[]``
+   containing range lengths) by mte_tags_to_ranges(). Note that
+   ``r_sizes[]`` sums up to 256 (``MTE_GRANULES_PER_PAGE``).
+
+   If ``r_sizes[]`` consists of a single element
+   (``{ MTE_GRANULES_PER_PAGE }``), the corresponding range is split into two
+   halves, i.e.::
+
+     r_sizes_new[2] = { MTE_GRANULES_PER_PAGE/2, MTE_GRANULES_PER_PAGE/2 };
+     r_tags_new[2] = { r_tags[0], r_tags[0] };
+
+2. 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 exceeds
+   ``MTE_GRANULES_PER_PAGE/2``.
+
+3. If the number ``N`` of ranges does not exceed ``6``, the ranges can be
+   compressed into 64 bits. This is done by storing the following values packed
+   into the pointer (``i<size>`` means a ``<size>``-bit unsigned integer)
+   treated as a bitmap (see ``include/linux/bitmap.h``)::
+
+    bit 0      :      (always 1) : i1
+    bits 1-3   :     largest_idx : i3
+    bits 4-27  :    r_tags[0..5] : i4 x 6
+    bits 28-62 :   r_sizes[0..4] : i7 x 5
+    bit 63     :      (always 0) : i1
+
+   If N is less than 6, ``r_tags`` and ``r_sizes`` are padded up with zero
+   values. The unused bits in the pointer, including bit 63, are also set to 0,
+   so the compressed data can be stored in XArray.
+
+   Range size of ``MTE_GRANULES_PER_PAGE/2`` (at most one) does not fit into
+   i7 and will be written as 0. This case is handled separately by the
+   decompressing procedure.
+
+Tag decompression
+-----------------
+
+The decompression algorithm performs the steps below.
+
+1. Read the lowest bit of the data from the input buffer and check that it is 1,
+   otherwise bail out.
+
+2. Read ``largest_idx``, ``r_tags[]`` and ``r_sizes[]`` from the
+   input buffer.
+
+   If ``largest_idx`` is zero, and all ``r_sizes[]`` are zero, set
+   ``r_sizes[0] = MTE_GRANULES_PER_PAGE/2``.
+
+   Calculate the removed largest element of ``r_sizes[]`` as
+   ``largest = 256 - sum(r_sizes)`` and insert it into ``r_sizes`` at
+   position ``largest_idx``.
+
+6. For each ``r_sizes[i] > 0``, add a 4-bit value ``r_tags[i]`` to the output
+   buffer ``r_sizes[i]`` times.
+
+
+Why these numbers?
+------------------
+
+To be able to reconstruct ``N`` tag ranges from the compressed data, we need to
+store the indicator bit together with ``largest_idx``, ``r_tags[N]``, and
+``r_sizes[N-1]`` in 63 bits.
+Knowing that the sizes do not exceed ``MTE_PAGE_TAG_STORAGE``, each of them can be
+packed into ``S = ilog2(MTE_PAGE_TAG_STORAGE)`` bits, whereas a single tag occupies
+4 bits.
+
+It is evident that the number of ranges that can be stored in 63 bits is
+strictly less than 8, therefore we only need 3 bits to store ``largest_idx``.
+
+The maximum values of ``N`` so that the number ``1 + 3 + N * 4 + (N-1) * S`` of
+storage bits does not exceed 63, are shown in the table below::
+
+ +-----------+-----------------+----+---+-------------------+
+ | Page size | Tag buffer size |  S | N |    Storage bits   |
+ +-----------+-----------------+----+---+-------------------+
+ |      4 KB |           128 B |  7 | 6 | 63 = 1+3+6*4+5*7  |
+ |     16 KB |           512 B |  9 | 5 | 60 = 1+3+5*4+4*9  |
+ |     64 KB |          2048 B | 11 | 4 | 53 = 1+3+4*4+3*11 |
+ +-----------+-----------------+----+---+-------------------+
+
+Note
+----
+
+Tag compression and decompression implicitly rely on the fixed MTE tag size
+(4 bits) and number of tags per page. Should these values change, the algorithm
+may need to be revised.
+
+
+Programming Interface
+=====================
+
+ .. kernel-doc:: arch/arm64/include/asm/mtecomp.h
+ .. kernel-doc:: arch/arm64/mm/mtecomp.c
+   :export:
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 7b071a00425d2..5f4d4b49a512e 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -2078,6 +2078,17 @@ 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 Memory Tagging Extension"
+	default y
+	depends on ARM64_MTE
+	help
+	  Enable tag compression support for ARM64 Memory Tagging Extension.
+
+	  Tag buffers corresponding to swapped RAM pages are compressed using
+	  RLE to conserve heap memory. In the common case compressed tags
+	  occupy 2.5x less 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..b9a3a921a38d4
--- /dev/null
+++ b/arch/arm64/include/asm/mtecomp.h
@@ -0,0 +1,39 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef __ASM_MTECOMP_H
+#define __ASM_MTECOMP_H
+
+#include <linux/types.h>
+
+/**
+ * mte_is_compressed() - check if the supplied pointer contains compressed tags.
+ * @ptr: pointer returned by kmalloc() or mte_compress().
+ *
+ * Returns: true iff bit 0 of @ptr is 1, which is only possible if @ptr was
+ * returned by mte_is_compressed().
+ */
+static inline bool mte_is_compressed(void *ptr)
+{
+	return ((unsigned long)ptr & 1);
+}
+
+#if defined(CONFIG_ARM64_MTE_COMP)
+
+void *mte_compress(u8 *tags);
+bool mte_decompress(void *handle, u8 *tags);
+
+#else
+
+static inline void *mte_compress(u8 *tags)
+{
+	return NULL;
+}
+
+static inline bool mte_decompress(void *data, u8 *tags)
+{
+	return false;
+}
+
+#endif // CONFIG_ARM64_MTE_COMP
+
+#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..c948921525030
--- /dev/null
+++ b/arch/arm64/mm/mtecomp.c
@@ -0,0 +1,257 @@
+// SPDX-License-Identifier: GPL-2.0-only
+
+/*
+ * MTE tag compression algorithm.
+ * See Documentation/arch/arm64/mte-tag-compression.rst for more details.
+ */
+
+#include <linux/bits.h>
+#include <linux/bitmap.h>
+#include <linux/bitops.h>
+#include <linux/export.h>
+#include <linux/gfp.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+#include <asm/mtecomp.h>
+
+#include "mtecomp.h"
+
+#define MTE_BITS_PER_LARGEST_IDX 3
+/* Range size cannot exceed MTE_GRANULES_PER_PAGE / 2. */
+#define MTE_BITS_PER_SIZE (ilog2(MTE_GRANULES_PER_PAGE) - 1)
+
+/*
+ * See Documentation/arch/arm64/mte-tag-compression.rst for details on how the
+ * maximum number of ranges is calculated.
+ */
+#if defined(CONFIG_ARM64_4K_PAGES)
+#define MTE_MAX_RANGES 6
+#elif defined(CONFIG_ARM64_16K_PAGES)
+#define MTE_MAX_RANGES 5
+#else
+#define MTE_MAX_RANGES 4
+#endif
+
+/**
+ * mte_tags_to_ranges() - break @tags into arrays of tag ranges.
+ * @tags: MTE_GRANULES_PER_PAGE-byte array containing MTE tags.
+ * @out_tags: u8 array to store the tag of every range.
+ * @out_sizes: unsigned short 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[]).
+ *
+ * This function is exported for testing purposes.
+ */
+void mte_tags_to_ranges(u8 *tags, u8 *out_tags, unsigned short *out_sizes,
+			size_t *out_len)
+{
+	u8 prev_tag = tags[0] / 16; /* First tag in the array. */
+	unsigned int cur_idx = 0, i, j;
+	u8 cur_tag;
+
+	memset(out_tags, 0, array_size(*out_len, sizeof(*out_tags)));
+	memset(out_sizes, 0, array_size(*out_len, sizeof(*out_sizes)));
+
+	out_tags[cur_idx] = prev_tag;
+	for (i = 0; i < MTE_GRANULES_PER_PAGE; i++) {
+		j = i % 2;
+		cur_tag = j ? (tags[i / 2] % 16) : (tags[i / 2] / 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(mte_tags_to_ranges, MTECOMP);
+
+/**
+ * mte_ranges_to_tags() - fill @tags using given tag ranges.
+ * @r_tags: u8[] containing the tag of every range.
+ * @r_sizes: unsigned short[] containing the size of every range.
+ * @r_len: length of @r_tags and @r_sizes.
+ * @tags: MTE_GRANULES_PER_PAGE-byte array to write the tags to.
+ *
+ * This function is exported for testing purposes.
+ */
+void mte_ranges_to_tags(u8 *r_tags, unsigned short *r_sizes, size_t r_len,
+			u8 *tags)
+{
+	unsigned 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(mte_ranges_to_tags, MTECOMP);
+
+static void mte_bitmap_write(unsigned long *bitmap, unsigned long value,
+			     unsigned long *pos, unsigned long bits)
+{
+	bitmap_write(bitmap, value, *pos, bits);
+	*pos += bits;
+}
+
+/* Compress ranges into an unsigned long. */
+static void mte_compress_to_ulong(size_t len, u8 *tags, unsigned short *sizes,
+				  unsigned long *result)
+{
+	unsigned long bit_pos = 0;
+	unsigned int largest_idx, i;
+	unsigned short largest = 0;
+
+	for (i = 0; i < len; i++) {
+		if (sizes[i] > largest) {
+			largest = sizes[i];
+			largest_idx = i;
+		}
+	}
+	/* Bit 1 in position 0 indicates compressed data. */
+	mte_bitmap_write(result, 1, &bit_pos, 1);
+	mte_bitmap_write(result, largest_idx, &bit_pos,
+			 MTE_BITS_PER_LARGEST_IDX);
+	for (i = 0; i < len; i++)
+		mte_bitmap_write(result, tags[i], &bit_pos, MTE_TAG_SIZE);
+	if (len == 1) {
+		/*
+		 * We are compressing MTE_GRANULES_PER_PAGE of identical tags.
+		 * Split it into two ranges containing
+		 * MTE_GRANULES_PER_PAGE / 2 tags, so that it falls into the
+		 * special case described below.
+		 */
+		mte_bitmap_write(result, tags[0], &bit_pos, MTE_TAG_SIZE);
+		i = 2;
+	} else {
+		i = len;
+	}
+	for (; i < MTE_MAX_RANGES; i++)
+		mte_bitmap_write(result, 0, &bit_pos, MTE_TAG_SIZE);
+	/*
+	 * Most of the time sizes[i] fits into MTE_BITS_PER_SIZE, apart from a
+	 * special case when:
+	 *   len = 2;
+	 *   sizes = { MTE_GRANULES_PER_PAGE / 2, MTE_GRANULES_PER_PAGE / 2};
+	 * In this case largest_idx will be set to 0, and the size written to
+	 * the bitmap will be also 0.
+	 */
+	for (i = 0; i < len; i++) {
+		if (i != largest_idx)
+			mte_bitmap_write(result, sizes[i], &bit_pos,
+					 MTE_BITS_PER_SIZE);
+	}
+	for (i = len; i < MTE_MAX_RANGES; i++)
+		mte_bitmap_write(result, 0, &bit_pos, MTE_BITS_PER_SIZE);
+}
+
+/**
+ * mte_compress() - compress the given tag array.
+ * @tags: MTE_GRANULES_PER_PAGE-byte array to read the tags from.
+ *
+ * Attempts to compress the user-supplied tag array.
+ *
+ * Returns: compressed data or NULL.
+ */
+void *mte_compress(u8 *tags)
+{
+	unsigned short *r_sizes;
+	void *result = NULL;
+	u8 *r_tags;
+	size_t r_len;
+
+	r_sizes = kmalloc_array(MTE_GRANULES_PER_PAGE, sizeof(unsigned short),
+				GFP_KERNEL);
+	r_tags = kmalloc(MTE_GRANULES_PER_PAGE, GFP_KERNEL);
+	if (!r_sizes || !r_tags)
+		goto ret;
+	r_len = MTE_GRANULES_PER_PAGE;
+	mte_tags_to_ranges(tags, r_tags, r_sizes, &r_len);
+	if (r_len <= MTE_MAX_RANGES)
+		mte_compress_to_ulong(r_len, r_tags, r_sizes,
+				      (unsigned long *)&result);
+ret:
+	kfree(r_tags);
+	kfree(r_sizes);
+	return result;
+}
+EXPORT_SYMBOL_NS(mte_compress, MTECOMP);
+
+static unsigned long mte_bitmap_read(const unsigned long *bitmap,
+				     unsigned long *pos, unsigned long bits)
+{
+	unsigned long start = *pos;
+
+	*pos += bits;
+	return bitmap_read(bitmap, start, bits);
+}
+
+/**
+ * mte_decompress() - decompress the tag array from the given pointer.
+ * @data: pointer returned by @mte_compress()
+ * @tags: MTE_GRANULES_PER_PAGE-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 if the passed data is uncompressed.
+ */
+bool mte_decompress(void *data, u8 *tags)
+{
+	unsigned short r_sizes[MTE_MAX_RANGES];
+	u8 r_tags[MTE_MAX_RANGES];
+	unsigned int largest_idx, i;
+	unsigned long bit_pos = 0;
+	unsigned long *bitmap;
+	unsigned short sum;
+	size_t max_ranges;
+
+	if (!mte_is_compressed(data))
+		return false;
+
+	bitmap = (unsigned long *)&data;
+	max_ranges = MTE_MAX_RANGES;
+	/* Skip the leading bit indicating the inline case. */
+	mte_bitmap_read(bitmap, &bit_pos, 1);
+	largest_idx =
+		mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_LARGEST_IDX);
+	if (largest_idx >= MTE_MAX_RANGES)
+		return false;
+
+	for (i = 0; i < max_ranges; i++)
+		r_tags[i] = mte_bitmap_read(bitmap, &bit_pos, MTE_TAG_SIZE);
+	for (i = 0, sum = 0; i < max_ranges; i++) {
+		if (i == largest_idx)
+			continue;
+		r_sizes[i] =
+			mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_SIZE);
+		/*
+		 * Special case: tag array consists of two ranges of
+		 * `MTE_GRANULES_PER_PAGE / 2` tags.
+		 */
+		if ((largest_idx == 0) && (i == 1) && (r_sizes[i] == 0))
+			r_sizes[i] = MTE_GRANULES_PER_PAGE / 2;
+		if (!r_sizes[i]) {
+			max_ranges = i;
+			break;
+		}
+		sum += r_sizes[i];
+	}
+	if (sum >= MTE_GRANULES_PER_PAGE)
+		return false;
+	r_sizes[largest_idx] = MTE_GRANULES_PER_PAGE - sum;
+	mte_ranges_to_tags(r_tags, r_sizes, max_ranges, tags);
+	return true;
+}
+EXPORT_SYMBOL_NS(mte_decompress, MTECOMP);
diff --git a/arch/arm64/mm/mtecomp.h b/arch/arm64/mm/mtecomp.h
new file mode 100644
index 0000000000000..b94cf0384f2af
--- /dev/null
+++ b/arch/arm64/mm/mtecomp.h
@@ -0,0 +1,12 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+
+#ifndef ARCH_ARM64_MM_MTECOMP_H_
+#define ARCH_ARM64_MM_MTECOMP_H_
+
+/* Functions exported from mtecomp.c for test_mtecomp.c. */
+void mte_tags_to_ranges(u8 *tags, u8 *out_tags, unsigned short *out_sizes,
+			size_t *out_len);
+void mte_ranges_to_tags(u8 *r_tags, unsigned short *r_sizes, size_t r_len,
+			u8 *tags);
+
+#endif  // ARCH_ARM64_MM_TEST_MTECOMP_H_
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Yury Norov]

On Thu, Dec 14, 2023 at 12:06:36PM +0100, Alexander Potapenko wrote:
> The config implements the algorithm compressing memory tags for ARM MTE
> during swapping.
> 
> The algorithm is based on RLE and specifically targets buffers of tags
> corresponding to a single page. In many cases a buffer can be compressed
> into 63 bits, making it possible to store it without additional memory
> allocation.
> 
> Suggested-by: Evgenii Stepanov <eugenis@google.com>
> Signed-off-by: Alexander Potapenko <glider@google.com>
> Acked-by: Catalin Marinas <catalin.marinas@arm.com>
> 
> ---
>  v10-mte:
>   - added Catalin's Acked-by:
> 
>  v8:
>   - As suggested by Catalin Marinas, only compress tags if they can be
>     stored inline. This simplifies the code drastically.
>   - Update the documentation.
>   - Split off patches introducing bitmap_read()/bitmap_write().
> 
>  v6:
>   - shuffle bits in inline handles so that they can't be confused with
>     canonical pointers;
>   - use kmem_cache_zalloc() to allocate compressed storage
>   - correctly handle range size overflow
>   - minor documentation fixes, clarify the special cases
> 
>  v5:
>   - make code PAGE_SIZE-agnostic, remove hardcoded constants, updated
>     the docs
>   - implement debugfs interface
>   - Address comments by Andy Shevchenko:
>     - update description of mtecomp.c
>     - remove redundant assignments, simplify mte_tags_to_ranges()
>     - various code simplifications
>     - introduce mtecomp.h
>     - add :export: to Documentation/arch/arm64/mte-tag-compression.rst
> 
>  v4:
>   - Addressed comments by Andy Shevchenko:
>     - expanded "MTE" to "Memory Tagging Extension" in Kconfig
>     - fixed kernel-doc comments, moved them to C source
>     - changed variables to unsigned where applicable
>     - some code simplifications, fewer unnecessary assignments
>     - added the mte_largest_idx_bits() helper
>     - added namespace prefixes to all functions
>     - added missing headers (but removed bits.h)
>   - Addressed comments by Yury Norov:
>     - removed test-only functions from mtecomp.h
>     - dropped the algoritm name (all functions are now prefixed with
>       "mte")
>     - added more comments
>     - got rid of MTE_RANGES_INLINE
>     - renamed bitmap_{get,set}_value() to bitmap_{read,write}()
>     - moved the big comment explaining the algorithm to
>       Documentation/arch/arm64/mte-tag-compression.rst, expanded it,
>       add a link to it from Documentation/arch/arm64/index.rst
>     - removed hardcoded ranges from mte_alloc_size()/mte_size_to_ranges()
> 
>  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 Yury Norov, switched from struct bitq (which is
>     not needed anymore) to <linux/bitmap.h>
>   - add missing symbol exports
> ---
>  Documentation/arch/arm64/index.rst            |   1 +
>  .../arch/arm64/mte-tag-compression.rst        | 154 +++++++++++
>  arch/arm64/Kconfig                            |  11 +
>  arch/arm64/include/asm/mtecomp.h              |  39 +++
>  arch/arm64/mm/Makefile                        |   1 +
>  arch/arm64/mm/mtecomp.c                       | 257 ++++++++++++++++++
>  arch/arm64/mm/mtecomp.h                       |  12 +
>  7 files changed, 475 insertions(+)
>  create mode 100644 Documentation/arch/arm64/mte-tag-compression.rst
>  create mode 100644 arch/arm64/include/asm/mtecomp.h
>  create mode 100644 arch/arm64/mm/mtecomp.c
>  create mode 100644 arch/arm64/mm/mtecomp.h
> 
> diff --git a/Documentation/arch/arm64/index.rst b/Documentation/arch/arm64/index.rst
> index d08e924204bf1..bf6c1583233a9 100644
> --- a/Documentation/arch/arm64/index.rst
> +++ b/Documentation/arch/arm64/index.rst
> @@ -19,6 +19,7 @@ ARM64 Architecture
>      legacy_instructions
>      memory
>      memory-tagging-extension
> +    mte-tag-compression
>      perf
>      pointer-authentication
>      ptdump
> diff --git a/Documentation/arch/arm64/mte-tag-compression.rst b/Documentation/arch/arm64/mte-tag-compression.rst
> new file mode 100644
> index 0000000000000..8fe6b51a9db6d
> --- /dev/null
> +++ b/Documentation/arch/arm64/mte-tag-compression.rst
> @@ -0,0 +1,154 @@
> +.. SPDX-License-Identifier: GPL-2.0
> +
> +==================================================
> +Tag Compression for Memory Tagging Extension (MTE)
> +==================================================
> +
> +This document describes the algorithm used to compress memory tags used by the
> +ARM Memory Tagging Extension (MTE).
> +
> +Introduction
> +============
> +
> +MTE assigns tags to memory pages: for 4K pages those tags occupy 128 bytes
> +(256 4-bit tags each corresponding to a 16-byte MTE granule), for 16K pages -
> +512 bytes, for 64K pages - 2048 bytes. By default, MTE carves out 3.125% (1/16)
> +of the available physical memory to store the tags.
> +
> +When MTE pages are saved to swap, their tags need to be stored in the kernel
> +memory. If the system swap is used heavily, these tags may take a substantial
> +portion of the physical memory. To reduce memory waste, ``CONFIG_ARM64_MTE_COMP``
> +allows the kernel to store the tags in compressed form.
> +
> +Implementation details
> +======================
> +
> +The algorithm attempts to compress an array of ``MTE_PAGE_TAG_STORAGE``
> +tag bytes into a byte sequence that can be stored in an 8-byte pointer. If that
> +is not possible, the data is stored uncompressed.
> +
> +Tag manipulation and storage
> +----------------------------
> +
> +Tags for swapped pages are stored in an XArray that maps swap entries to 63-bit
> +values (see ``arch/arm64/mm/mteswap.c``). Bit 0 of these values indicates how
> +their contents should be treated:
> +
> + - 0: value is a pointer to an uncompressed buffer allocated with kmalloc()
> +   (always the case if ``CONFIG_ARM64_MTE_COMP=n``) with the highest bit set
> +   to 0;
> + - 1: value contains compressed data.
> +
> +``arch/arm64/include/asm/mtecomp.h`` declares the following functions that
> +manipulate with tags:
> +
> +- mte_compress() - compresses the given ``MTE_PAGE_TAG_STORAGE``-byte ``tags``
> +  buffer into a pointer;
> +- mte_decompress() - decompresses the tags from a pointer;
> +- mte_is_compressed() - returns ``true`` iff the pointer passed to it should be
> +  treated as compressed data.
> +
> +Tag compression
> +---------------
> +
> +The compression algorithm is a variation of RLE (run-length encoding) and works
> +as follows (we will be considering 4K pages and 128-byte tag buffers, but the
> +same approach scales to 16K and 64K pages):
> +
> +1. The input array of 128 (``MTE_PAGE_TAG_STORAGE``) bytes is transformed into
> +   tag ranges (two arrays: ``r_tags[]`` containing tag values and ``r_sizes[]``
> +   containing range lengths) by mte_tags_to_ranges(). Note that
> +   ``r_sizes[]`` sums up to 256 (``MTE_GRANULES_PER_PAGE``).
> +
> +   If ``r_sizes[]`` consists of a single element
> +   (``{ MTE_GRANULES_PER_PAGE }``), the corresponding range is split into two
> +   halves, i.e.::
> +
> +     r_sizes_new[2] = { MTE_GRANULES_PER_PAGE/2, MTE_GRANULES_PER_PAGE/2 };
> +     r_tags_new[2] = { r_tags[0], r_tags[0] };
> +
> +2. 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 exceeds
> +   ``MTE_GRANULES_PER_PAGE/2``.
> +
> +3. If the number ``N`` of ranges does not exceed ``6``, the ranges can be
> +   compressed into 64 bits. This is done by storing the following values packed
> +   into the pointer (``i<size>`` means a ``<size>``-bit unsigned integer)
> +   treated as a bitmap (see ``include/linux/bitmap.h``)::
> +
> +    bit 0      :      (always 1) : i1
> +    bits 1-3   :     largest_idx : i3
> +    bits 4-27  :    r_tags[0..5] : i4 x 6
> +    bits 28-62 :   r_sizes[0..4] : i7 x 5
> +    bit 63     :      (always 0) : i1
> +
> +   If N is less than 6, ``r_tags`` and ``r_sizes`` are padded up with zero
> +   values. The unused bits in the pointer, including bit 63, are also set to 0,
> +   so the compressed data can be stored in XArray.
> +
> +   Range size of ``MTE_GRANULES_PER_PAGE/2`` (at most one) does not fit into
> +   i7 and will be written as 0. This case is handled separately by the
> +   decompressing procedure.
> +
> +Tag decompression
> +-----------------
> +
> +The decompression algorithm performs the steps below.
> +
> +1. Read the lowest bit of the data from the input buffer and check that it is 1,
> +   otherwise bail out.
> +
> +2. Read ``largest_idx``, ``r_tags[]`` and ``r_sizes[]`` from the
> +   input buffer.
> +
> +   If ``largest_idx`` is zero, and all ``r_sizes[]`` are zero, set
> +   ``r_sizes[0] = MTE_GRANULES_PER_PAGE/2``.
> +
> +   Calculate the removed largest element of ``r_sizes[]`` as
> +   ``largest = 256 - sum(r_sizes)`` and insert it into ``r_sizes`` at
> +   position ``largest_idx``.
> +
> +6. For each ``r_sizes[i] > 0``, add a 4-bit value ``r_tags[i]`` to the output
> +   buffer ``r_sizes[i]`` times.
> +
> +
> +Why these numbers?
> +------------------
> +
> +To be able to reconstruct ``N`` tag ranges from the compressed data, we need to
> +store the indicator bit together with ``largest_idx``, ``r_tags[N]``, and
> +``r_sizes[N-1]`` in 63 bits.
> +Knowing that the sizes do not exceed ``MTE_PAGE_TAG_STORAGE``, each of them can be
> +packed into ``S = ilog2(MTE_PAGE_TAG_STORAGE)`` bits, whereas a single tag occupies
> +4 bits.
> +
> +It is evident that the number of ranges that can be stored in 63 bits is
> +strictly less than 8, therefore we only need 3 bits to store ``largest_idx``.
> +
> +The maximum values of ``N`` so that the number ``1 + 3 + N * 4 + (N-1) * S`` of
> +storage bits does not exceed 63, are shown in the table below::
> +
> + +-----------+-----------------+----+---+-------------------+
> + | Page size | Tag buffer size |  S | N |    Storage bits   |
> + +-----------+-----------------+----+---+-------------------+
> + |      4 KB |           128 B |  7 | 6 | 63 = 1+3+6*4+5*7  |
> + |     16 KB |           512 B |  9 | 5 | 60 = 1+3+5*4+4*9  |
> + |     64 KB |          2048 B | 11 | 4 | 53 = 1+3+4*4+3*11 |
> + +-----------+-----------------+----+---+-------------------+
> +
> +Note
> +----
> +
> +Tag compression and decompression implicitly rely on the fixed MTE tag size
> +(4 bits) and number of tags per page. Should these values change, the algorithm
> +may need to be revised.
> +
> +
> +Programming Interface
> +=====================
> +
> + .. kernel-doc:: arch/arm64/include/asm/mtecomp.h
> + .. kernel-doc:: arch/arm64/mm/mtecomp.c
> +   :export:
> diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
> index 7b071a00425d2..5f4d4b49a512e 100644
> --- a/arch/arm64/Kconfig
> +++ b/arch/arm64/Kconfig
> @@ -2078,6 +2078,17 @@ 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 Memory Tagging Extension"
> +	default y
> +	depends on ARM64_MTE
> +	help
> +	  Enable tag compression support for ARM64 Memory Tagging Extension.
> +
> +	  Tag buffers corresponding to swapped RAM pages are compressed using
> +	  RLE to conserve heap memory. In the common case compressed tags
> +	  occupy 2.5x less 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..b9a3a921a38d4
> --- /dev/null
> +++ b/arch/arm64/include/asm/mtecomp.h
> @@ -0,0 +1,39 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef __ASM_MTECOMP_H
> +#define __ASM_MTECOMP_H
> +
> +#include <linux/types.h>
> +
> +/**
> + * mte_is_compressed() - check if the supplied pointer contains compressed tags.
> + * @ptr: pointer returned by kmalloc() or mte_compress().
> + *
> + * Returns: true iff bit 0 of @ptr is 1, which is only possible if @ptr was
> + * returned by mte_is_compressed().
> + */
> +static inline bool mte_is_compressed(void *ptr)
> +{
> +	return ((unsigned long)ptr & 1);
> +}
> +
> +#if defined(CONFIG_ARM64_MTE_COMP)
> +
> +void *mte_compress(u8 *tags);
> +bool mte_decompress(void *handle, u8 *tags);
> +
> +#else
> +
> +static inline void *mte_compress(u8 *tags)
> +{
> +	return NULL;
> +}
> +
> +static inline bool mte_decompress(void *data, u8 *tags)
> +{
> +	return false;
> +}
> +
> +#endif // CONFIG_ARM64_MTE_COMP
> +
> +#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..c948921525030
> --- /dev/null
> +++ b/arch/arm64/mm/mtecomp.c
> @@ -0,0 +1,257 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +
> +/*
> + * MTE tag compression algorithm.
> + * See Documentation/arch/arm64/mte-tag-compression.rst for more details.
> + */
> +
> +#include <linux/bits.h>
> +#include <linux/bitmap.h>
> +#include <linux/bitops.h>
> +#include <linux/export.h>
> +#include <linux/gfp.h>
> +#include <linux/module.h>
> +#include <linux/slab.h>
> +#include <linux/string.h>
> +#include <linux/types.h>
> +
> +#include <asm/mtecomp.h>
> +
> +#include "mtecomp.h"
> +
> +#define MTE_BITS_PER_LARGEST_IDX 3
> +/* Range size cannot exceed MTE_GRANULES_PER_PAGE / 2. */
> +#define MTE_BITS_PER_SIZE (ilog2(MTE_GRANULES_PER_PAGE) - 1)
> +
> +/*
> + * See Documentation/arch/arm64/mte-tag-compression.rst for details on how the
> + * maximum number of ranges is calculated.
> + */
> +#if defined(CONFIG_ARM64_4K_PAGES)
> +#define MTE_MAX_RANGES 6
> +#elif defined(CONFIG_ARM64_16K_PAGES)
> +#define MTE_MAX_RANGES 5
> +#else
> +#define MTE_MAX_RANGES 4
> +#endif
> +
> +/**
> + * mte_tags_to_ranges() - break @tags into arrays of tag ranges.
> + * @tags: MTE_GRANULES_PER_PAGE-byte array containing MTE tags.
> + * @out_tags: u8 array to store the tag of every range.
> + * @out_sizes: unsigned short 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[]).
> + *
> + * This function is exported for testing purposes.
> + */
> +void mte_tags_to_ranges(u8 *tags, u8 *out_tags, unsigned short *out_sizes,
> +			size_t *out_len)
> +{
> +	u8 prev_tag = tags[0] / 16; /* First tag in the array. */
> +	unsigned int cur_idx = 0, i, j;
> +	u8 cur_tag;
> +
> +	memset(out_tags, 0, array_size(*out_len, sizeof(*out_tags)));
> +	memset(out_sizes, 0, array_size(*out_len, sizeof(*out_sizes)));
> +
> +	out_tags[cur_idx] = prev_tag;
> +	for (i = 0; i < MTE_GRANULES_PER_PAGE; i++) {
> +		j = i % 2;
> +		cur_tag = j ? (tags[i / 2] % 16) : (tags[i / 2] / 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(mte_tags_to_ranges, MTECOMP);
> +
> +/**
> + * mte_ranges_to_tags() - fill @tags using given tag ranges.
> + * @r_tags: u8[] containing the tag of every range.
> + * @r_sizes: unsigned short[] containing the size of every range.
> + * @r_len: length of @r_tags and @r_sizes.
> + * @tags: MTE_GRANULES_PER_PAGE-byte array to write the tags to.
> + *
> + * This function is exported for testing purposes.
> + */
> +void mte_ranges_to_tags(u8 *r_tags, unsigned short *r_sizes, size_t r_len,
> +			u8 *tags)
> +{
> +	unsigned 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(mte_ranges_to_tags, MTECOMP);
> +
> +static void mte_bitmap_write(unsigned long *bitmap, unsigned long value,
> +			     unsigned long *pos, unsigned long bits)
> +{
> +	bitmap_write(bitmap, value, *pos, bits);
> +	*pos += bits;
> +}
> +
> +/* Compress ranges into an unsigned long. */
> +static void mte_compress_to_ulong(size_t len, u8 *tags, unsigned short *sizes,
> +				  unsigned long *result)
> +{
> +	unsigned long bit_pos = 0;
> +	unsigned int largest_idx, i;
> +	unsigned short largest = 0;
> +
> +	for (i = 0; i < len; i++) {
> +		if (sizes[i] > largest) {
> +			largest = sizes[i];
> +			largest_idx = i;
> +		}
> +	}
> +	/* Bit 1 in position 0 indicates compressed data. */
> +	mte_bitmap_write(result, 1, &bit_pos, 1);
> +	mte_bitmap_write(result, largest_idx, &bit_pos,
> +			 MTE_BITS_PER_LARGEST_IDX);
> +	for (i = 0; i < len; i++)
> +		mte_bitmap_write(result, tags[i], &bit_pos, MTE_TAG_SIZE);
> +	if (len == 1) {
> +		/*
> +		 * We are compressing MTE_GRANULES_PER_PAGE of identical tags.
> +		 * Split it into two ranges containing
> +		 * MTE_GRANULES_PER_PAGE / 2 tags, so that it falls into the
> +		 * special case described below.
> +		 */
> +		mte_bitmap_write(result, tags[0], &bit_pos, MTE_TAG_SIZE);
> +		i = 2;
> +	} else {
> +		i = len;
> +	}
> +	for (; i < MTE_MAX_RANGES; i++)
> +		mte_bitmap_write(result, 0, &bit_pos, MTE_TAG_SIZE);
> +	/*
> +	 * Most of the time sizes[i] fits into MTE_BITS_PER_SIZE, apart from a
> +	 * special case when:
> +	 *   len = 2;
> +	 *   sizes = { MTE_GRANULES_PER_PAGE / 2, MTE_GRANULES_PER_PAGE / 2};
> +	 * In this case largest_idx will be set to 0, and the size written to
> +	 * the bitmap will be also 0.
> +	 */
> +	for (i = 0; i < len; i++) {
> +		if (i != largest_idx)
> +			mte_bitmap_write(result, sizes[i], &bit_pos,
> +					 MTE_BITS_PER_SIZE);
> +	}
> +	for (i = len; i < MTE_MAX_RANGES; i++)
> +		mte_bitmap_write(result, 0, &bit_pos, MTE_BITS_PER_SIZE);
> +}
> +
> +/**
> + * mte_compress() - compress the given tag array.
> + * @tags: MTE_GRANULES_PER_PAGE-byte array to read the tags from.
> + *
> + * Attempts to compress the user-supplied tag array.
> + *
> + * Returns: compressed data or NULL.
> + */
> +void *mte_compress(u8 *tags)
> +{
> +	unsigned short *r_sizes;
> +	void *result = NULL;
> +	u8 *r_tags;
> +	size_t r_len;
> +
> +	r_sizes = kmalloc_array(MTE_GRANULES_PER_PAGE, sizeof(unsigned short),
> +				GFP_KERNEL);
> +	r_tags = kmalloc(MTE_GRANULES_PER_PAGE, GFP_KERNEL);
> +	if (!r_sizes || !r_tags)
> +		goto ret;
> +	r_len = MTE_GRANULES_PER_PAGE;
> +	mte_tags_to_ranges(tags, r_tags, r_sizes, &r_len);
> +	if (r_len <= MTE_MAX_RANGES)
> +		mte_compress_to_ulong(r_len, r_tags, r_sizes,
> +				      (unsigned long *)&result);
> +ret:
> +	kfree(r_tags);
> +	kfree(r_sizes);
> +	return result;
> +}
> +EXPORT_SYMBOL_NS(mte_compress, MTECOMP);
> +
> +static unsigned long mte_bitmap_read(const unsigned long *bitmap,
> +				     unsigned long *pos, unsigned long bits)
> +{
> +	unsigned long start = *pos;
> +
> +	*pos += bits;
> +	return bitmap_read(bitmap, start, bits);
> +}
> +
> +/**
> + * mte_decompress() - decompress the tag array from the given pointer.
> + * @data: pointer returned by @mte_compress()
> + * @tags: MTE_GRANULES_PER_PAGE-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 if the passed data is uncompressed.
> + */
> +bool mte_decompress(void *data, u8 *tags)
> +{
> +	unsigned short r_sizes[MTE_MAX_RANGES];
> +	u8 r_tags[MTE_MAX_RANGES];
> +	unsigned int largest_idx, i;
> +	unsigned long bit_pos = 0;
> +	unsigned long *bitmap;
> +	unsigned short sum;
> +	size_t max_ranges;
> +
> +	if (!mte_is_compressed(data))
> +		return false;
> +
> +	bitmap = (unsigned long *)&data;

That looks weird... You're casting address of a 'data' to a bitmap
instead of 'data'. At the 1st glance it makes little sense because
'data' is passed as parameter. Moreover, in mte_is_compressed()
you pass 'data', not '&data'. Can you please comment on your
intention?

> +	max_ranges = MTE_MAX_RANGES;
> +	/* Skip the leading bit indicating the inline case. */
> +	mte_bitmap_read(bitmap, &bit_pos, 1);
> +	largest_idx =
> +		mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_LARGEST_IDX);

Nit: really no need to split the line - we're OK with 100-chars per
line now.

> +	if (largest_idx >= MTE_MAX_RANGES)
> +		return false;
> +
> +	for (i = 0; i < max_ranges; i++)
> +		r_tags[i] = mte_bitmap_read(bitmap, &bit_pos, MTE_TAG_SIZE);
> +	for (i = 0, sum = 0; i < max_ranges; i++) {
> +		if (i == largest_idx)
> +			continue;
> +		r_sizes[i] =
> +			mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_SIZE);
> +		/*
> +		 * Special case: tag array consists of two ranges of
> +		 * `MTE_GRANULES_PER_PAGE / 2` tags.
> +		 */
> +		if ((largest_idx == 0) && (i == 1) && (r_sizes[i] == 0))
> +			r_sizes[i] = MTE_GRANULES_PER_PAGE / 2;
> +		if (!r_sizes[i]) {
> +			max_ranges = i;
> +			break;
> +		}
> +		sum += r_sizes[i];
> +	}
> +	if (sum >= MTE_GRANULES_PER_PAGE)
> +		return false;
> +	r_sizes[largest_idx] = MTE_GRANULES_PER_PAGE - sum;
> +	mte_ranges_to_tags(r_tags, r_sizes, max_ranges, tags);
> +	return true;
> +}
> +EXPORT_SYMBOL_NS(mte_decompress, MTECOMP);
> diff --git a/arch/arm64/mm/mtecomp.h b/arch/arm64/mm/mtecomp.h
> new file mode 100644
> index 0000000000000..b94cf0384f2af
> --- /dev/null
> +++ b/arch/arm64/mm/mtecomp.h
> @@ -0,0 +1,12 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +
> +#ifndef ARCH_ARM64_MM_MTECOMP_H_
> +#define ARCH_ARM64_MM_MTECOMP_H_
> +
> +/* Functions exported from mtecomp.c for test_mtecomp.c. */
> +void mte_tags_to_ranges(u8 *tags, u8 *out_tags, unsigned short *out_sizes,
> +			size_t *out_len);
> +void mte_ranges_to_tags(u8 *r_tags, unsigned short *r_sizes, size_t r_len,
> +			u8 *tags);
> +
> +#endif  // ARCH_ARM64_MM_TEST_MTECOMP_H_
> -- 
> 2.43.0.472.g3155946c3a-goog

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Alexander Potapenko]

>
> That looks weird... You're casting address of a 'data' to a bitmap
> instead of 'data'. At the 1st glance it makes little sense because
> 'data' is passed as parameter. Moreover, in mte_is_compressed()
> you pass 'data', not '&data'. Can you please comment on your
> intention?

Although `data` is a void*, it actually contains 64 bits of compressed
data, so we pass &data to mte_bitmap_read() to read its contents.
Perhaps I'd better make `data` an unsigned long to avoid confusion.

>
> > +     max_ranges = MTE_MAX_RANGES;
> > +     /* Skip the leading bit indicating the inline case. */
> > +     mte_bitmap_read(bitmap, &bit_pos, 1);
> > +     largest_idx =
> > +             mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_LARGEST_IDX);
>
> Nit: really no need to split the line - we're OK with 100-chars per
> line now.

That's true, but I am relying on clang-format here (maybe we should
extend the limit in .clang-format?)

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Yury Norov]

On Fri, Dec 15, 2023 at 04:19:27PM +0100, Alexander Potapenko wrote:
> >
> > That looks weird... You're casting address of a 'data' to a bitmap
> > instead of 'data'. At the 1st glance it makes little sense because
> > 'data' is passed as parameter. Moreover, in mte_is_compressed()
> > you pass 'data', not '&data'. Can you please comment on your
> > intention?
> 
> Although `data` is a void*, it actually contains 64 bits of compressed
> data, so we pass &data to mte_bitmap_read() to read its contents.
> Perhaps I'd better make `data` an unsigned long to avoid confusion.
 
Still don't understand. Let's consider this example:

yury:linux$ cat tst.c
#include <stdio.h>

unsigned long data[1] = {0xabc};

void foo(unsigned long *data)
{
	printf("foo:  *data\t%lx\n", (unsigned long)*data);
	printf("foo:   data\t%lx\n",  (unsigned long)data);
	printf("foo:  &data\t%lx\n", (unsigned long)&data);
}

void bar(unsigned long *data)
{
	volatile unsigned long x[100];

	printf("bar:  *data\t%lx\n", (unsigned long)*data);
	printf("bar:   data\t%lx\n",  (unsigned long)data);
	printf("bar:  &data\t%lx\n", (unsigned long)&data);
}

int main(int argc, char *argv[])
{

	foo(data);
	bar(data);

	printf("main: *data\t%lx\n", (unsigned long)*data);
	printf("main:  data\t%lx\n",  (unsigned long)data);
	printf("main: &data\t%lx\n", (unsigned long)&data);

	return 0;
}
yury:linux$ gcc tst.c -O0
yury:linux$ ./a.out
foo:  *data	abc
foo:   data	555b2cef9010
foo:  &data	7fff39d6e5f8
bar:  *data	abc
bar:   data	555b2cef9010
bar:  &data	7fff39d6e2c8
main: *data	abc
main:  data	555b2cef9010
main: &data	555b2cef9010

Data and *data have their meaning across scope boundary: a pointer and
a content. The &data is pretty much a random number - a pointer to
somewhere on a function's stack. Isn't?

> > > +     max_ranges = MTE_MAX_RANGES;
> > > +     /* Skip the leading bit indicating the inline case. */
> > > +     mte_bitmap_read(bitmap, &bit_pos, 1);
> > > +     largest_idx =
> > > +             mte_bitmap_read(bitmap, &bit_pos, MTE_BITS_PER_LARGEST_IDX);
> >
> > Nit: really no need to split the line - we're OK with 100-chars per
> > line now.
> 
> That's true, but I am relying on clang-format here (maybe we should
> extend the limit in .clang-format?)

If clang-format hurts readability, don't use it.

Thanks,
Yury

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Yury Norov]

On Fri, Dec 15, 2023 at 08:35:43AM -0800, Yury Norov wrote:
> On Fri, Dec 15, 2023 at 04:19:27PM +0100, Alexander Potapenko wrote:
> > >
> > > That looks weird... You're casting address of a 'data' to a bitmap
> > > instead of 'data'. At the 1st glance it makes little sense because
> > > 'data' is passed as parameter. Moreover, in mte_is_compressed()
> > > you pass 'data', not '&data'. Can you please comment on your
> > > intention?
> > 
> > Although `data` is a void*, it actually contains 64 bits of compressed
> > data, so we pass &data to mte_bitmap_read() to read its contents.
> > Perhaps I'd better make `data` an unsigned long to avoid confusion.
>  
> Still don't understand. Let's consider this example:
> 
> yury:linux$ cat tst.c
> #include <stdio.h>
> 
> unsigned long data[1] = {0xabc};
> 
> void foo(unsigned long *data)
> {
> 	printf("foo:  *data\t%lx\n", (unsigned long)*data);
> 	printf("foo:   data\t%lx\n",  (unsigned long)data);
> 	printf("foo:  &data\t%lx\n", (unsigned long)&data);
> }
> 
> void bar(unsigned long *data)
> {
> 	volatile unsigned long x[100];
> 
> 	printf("bar:  *data\t%lx\n", (unsigned long)*data);
> 	printf("bar:   data\t%lx\n",  (unsigned long)data);
> 	printf("bar:  &data\t%lx\n", (unsigned long)&data);
> }
> 
> int main(int argc, char *argv[])
> {
> 
> 	foo(data);
> 	bar(data);
> 
> 	printf("main: *data\t%lx\n", (unsigned long)*data);
> 	printf("main:  data\t%lx\n",  (unsigned long)data);
> 	printf("main: &data\t%lx\n", (unsigned long)&data);
> 
> 	return 0;
> }
> yury:linux$ gcc tst.c -O0
> yury:linux$ ./a.out
> foo:  *data	abc
> foo:   data	555b2cef9010
> foo:  &data	7fff39d6e5f8
> bar:  *data	abc
> bar:   data	555b2cef9010
> bar:  &data	7fff39d6e2c8
> main: *data	abc
> main:  data	555b2cef9010
> main: &data	555b2cef9010
> 
> Data and *data have their meaning across scope boundary: a pointer and
> a content. The &data is pretty much a random number - a pointer to
> somewhere on a function's stack. Isn't?

OK, I read now. Your 'void *data' is not a pointer, but an actual data.
Very confusing indeed... Maybe add a comment for that cast?
 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Alexander Potapenko]

> > Data and *data have their meaning across scope boundary: a pointer and
> > a content. The &data is pretty much a random number - a pointer to
> > somewhere on a function's stack. Isn't?
>
> OK, I read now. Your 'void *data' is not a pointer, but an actual data.
> Very confusing indeed... Maybe add a comment for that cast?
>
Sure, will do.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 4/7] arm64: mte: implement CONFIG_ARM64_MTE_COMP

[Alexander Potapenko]

> > > Nit: really no need to split the line - we're OK with 100-chars per
> > > line now.
> >
> > That's true, but I am relying on clang-format here (maybe we should
> > extend the limit in .clang-format?)
>
> If clang-format hurts readability, don't use it.
>
There's an old discussion on clang-format column limit being
inconsistent with what checkpatch mandates:
https://lore.kernel.org/lkml/20200610125147.2782142-1-christian.brauner@ubuntu.com/
I changed the name of the variable in this case to avoid hitting the
column limit.

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 5/7] arm64: mte: add a test for MTE tags compression

[Alexander Potapenko]

Ensure that tag sequences containing alternating values are compressed
to buffers of expected size and correctly decompressed afterwards.

Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

---
 v10-mte:
  - added Catalin's Acked-by:

 v9:
  - minor changes to Kconfig description

 v8:
  - adapt to the simplified compression algorithm

 v6:
  - add test_decompress_invalid() to ensure invalid handles are ignored;
  - add test_upper_bits(), which is a regression test for a case where
    an inline handle looked like an out-of-line one;
  - add test_compress_nonzero() to ensure a full nonzero tag array is
    compressed correctly;
  - add test_two_ranges() to test cases when the input buffer is divided
    into two ranges.

 v5:
  - remove hardcoded constants, added test setup/teardown;
  - support 16- and 64K pages;
  - replace nested if-clauses with expected_size_from_ranges();
  - call mte_release_handle() after tests that perform
    compression/decompression;
  - address comments by Andy Shevchenko:
    - fix include order;
    - use mtecomp.h instead of function prototypes.

 v4:
  - addressed comments by Andy Shevchenko:
    - expanded MTE to "Memory Tagging Extension" in Kconfig
    - changed signed variables to unsigned where applicable
    - added missing header dependencies

  - addressed comments by Yury Norov:
    - moved test-only declarations from mtecomp.h into this test
    - switched to the new "mte"-prefixed function names, dropped the
      mentions of "EA0"
    - added test_tag_to_ranges_n()

 v3:
  - addressed comments by Andy Shevchenko in another patch:
   - switched from u64 to unsigned long
   - added MODULE_IMPORT_NS(MTECOMP)
   - fixed includes order
---
 arch/arm64/Kconfig           |  11 ++
 arch/arm64/mm/Makefile       |   1 +
 arch/arm64/mm/test_mtecomp.c | 364 +++++++++++++++++++++++++++++++++++
 3 files changed, 376 insertions(+)
 create mode 100644 arch/arm64/mm/test_mtecomp.c

diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 5f4d4b49a512e..6a1397a96f2f0 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -2089,6 +2089,17 @@ config ARM64_MTE_COMP
 	  RLE to conserve heap memory. In the common case compressed tags
 	  occupy 2.5x less memory.
 
+config ARM64_MTE_COMP_KUNIT_TEST
+	tristate "Test tag compression for ARM64 Memory Tagging Extension" if !KUNIT_ALL_TESTS
+	default KUNIT_ALL_TESTS
+	depends on KUNIT && ARM64_MTE_COMP
+	help
+	  Test MTE compression algorithm enabled by CONFIG_ARM64_MTE_COMP.
+
+	  Ensure that certain tag sequences containing alternating values can
+	  be compressed into pointer-size values and correctly decompressed
+	  afterwards.
+
 config ARM64_SVE
 	bool "ARM Scalable Vector Extension support"
 	default y
diff --git a/arch/arm64/mm/Makefile b/arch/arm64/mm/Makefile
index 46778f6dd83c2..170dc62b010b9 100644
--- a/arch/arm64/mm/Makefile
+++ b/arch/arm64/mm/Makefile
@@ -11,6 +11,7 @@ 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
+obj-$(CONFIG_ARM64_MTE_COMP_KUNIT_TEST) += test_mtecomp.o
 KASAN_SANITIZE_physaddr.o	+= n
 
 obj-$(CONFIG_KASAN)		+= kasan_init.o
diff --git a/arch/arm64/mm/test_mtecomp.c b/arch/arm64/mm/test_mtecomp.c
new file mode 100644
index 0000000000000..e8aeb7607ff41
--- /dev/null
+++ b/arch/arm64/mm/test_mtecomp.c
@@ -0,0 +1,364 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Test cases for MTE tags compression algorithm.
+ */
+
+#include <linux/bits.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+#include <kunit/test.h>
+
+#include <asm/mtecomp.h>
+
+#include "mtecomp.h"
+
+/* Per-test storage allocated in mtecomp_test_init(). */
+struct test_data {
+	u8 *tags, *dtags;
+	unsigned short *r_sizes;
+	size_t r_len;
+	u8 *r_tags;
+};
+
+/*
+ * Split td->tags to ranges stored in td->r_tags, td->r_sizes, td->r_len,
+ * then convert those ranges back to tags stored in td->dtags.
+ */
+static void tags_to_ranges_to_tags_helper(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+
+	mte_tags_to_ranges(td->tags, td->r_tags, td->r_sizes, &td->r_len);
+	mte_ranges_to_tags(td->r_tags, td->r_sizes, td->r_len, td->dtags);
+	KUNIT_EXPECT_EQ(test, memcmp(td->tags, td->dtags, MTE_PAGE_TAG_STORAGE),
+			0);
+}
+
+/*
+ * Test that mte_tags_to_ranges() produces a single range for a zero-filled tag
+ * buffer.
+ */
+static void test_tags_to_ranges_zero(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+
+	memset(td->tags, 0, MTE_PAGE_TAG_STORAGE);
+	tags_to_ranges_to_tags_helper(test);
+
+	KUNIT_EXPECT_EQ(test, td->r_len, 1);
+	KUNIT_EXPECT_EQ(test, td->r_tags[0], 0);
+	KUNIT_EXPECT_EQ(test, td->r_sizes[0], MTE_GRANULES_PER_PAGE);
+}
+
+/*
+ * Test that a small number of different tags is correctly transformed into
+ * ranges.
+ */
+static void test_tags_to_ranges_simple(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	const u8 ex_tags[] = { 0xa, 0x0, 0xa, 0xb, 0x0 };
+	const unsigned short ex_sizes[] = { 1, 2, 2, 1,
+					    MTE_GRANULES_PER_PAGE - 6 };
+
+	memset(td->tags, 0, MTE_PAGE_TAG_STORAGE);
+	td->tags[0] = 0xa0;
+	td->tags[1] = 0x0a;
+	td->tags[2] = 0xab;
+	tags_to_ranges_to_tags_helper(test);
+
+	KUNIT_EXPECT_EQ(test, td->r_len, 5);
+	KUNIT_EXPECT_EQ(test, memcmp(td->r_tags, ex_tags, sizeof(ex_tags)), 0);
+	KUNIT_EXPECT_EQ(test, memcmp(td->r_sizes, ex_sizes, sizeof(ex_sizes)),
+			0);
+}
+
+/* Test that repeated 0xa0 byte produces MTE_GRANULES_PER_PAGE ranges of length 1. */
+static void test_tags_to_ranges_repeated(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+
+	memset(td->tags, 0xa0, MTE_PAGE_TAG_STORAGE);
+	tags_to_ranges_to_tags_helper(test);
+
+	KUNIT_EXPECT_EQ(test, td->r_len, MTE_GRANULES_PER_PAGE);
+}
+
+/* Generate a buffer that will contain @nranges of tag ranges. */
+static void gen_tag_range_helper(u8 *tags, int nranges)
+{
+	unsigned int i;
+
+	memset(tags, 0, MTE_PAGE_TAG_STORAGE);
+	if (nranges > 1) {
+		nranges--;
+		for (i = 0; i < nranges / 2; i++)
+			tags[i] = 0xab;
+		if (nranges % 2)
+			tags[nranges / 2] = 0xa0;
+	}
+}
+
+/*
+ * Test that mte_tags_to_ranges()/mte_ranges_to_tags() work for various
+ * r_len values.
+ */
+static void test_tag_to_ranges_n(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	unsigned int i, j, sum;
+
+	for (i = 1; i <= MTE_GRANULES_PER_PAGE; i++) {
+		gen_tag_range_helper(td->tags, i);
+		tags_to_ranges_to_tags_helper(test);
+		sum = 0;
+		for (j = 0; j < td->r_len; j++)
+			sum += td->r_sizes[j];
+		KUNIT_EXPECT_EQ(test, sum, MTE_GRANULES_PER_PAGE);
+	}
+}
+
+/*
+ * Check that the tag buffer in test->priv can be compressed and decompressed
+ * without changes.
+ */
+static void *compress_decompress_helper(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+
+	handle = mte_compress(td->tags);
+	KUNIT_EXPECT_EQ(test, (unsigned long)handle & BIT_ULL(63), 0);
+	if (handle) {
+		KUNIT_EXPECT_TRUE(test, mte_decompress(handle, td->dtags));
+		KUNIT_EXPECT_EQ(test, memcmp(td->tags, td->dtags, MTE_PAGE_TAG_STORAGE),
+				0);
+	}
+	return handle;
+}
+
+/* Test that a zero-filled array is compressed into inline storage. */
+static void test_compress_zero(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+
+	memset(td->tags, 0, MTE_PAGE_TAG_STORAGE);
+	handle = compress_decompress_helper(test);
+	/* Tags are stored inline. */
+	KUNIT_EXPECT_TRUE(test, mte_is_compressed(handle));
+}
+
+/* Test that a 0xaa-filled array is compressed into inline storage. */
+static void test_compress_nonzero(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+
+	memset(td->tags, 0xaa, MTE_PAGE_TAG_STORAGE);
+	handle = compress_decompress_helper(test);
+	/* Tags are stored inline. */
+	KUNIT_EXPECT_TRUE(test, mte_is_compressed(handle));
+}
+
+/*
+ * Test that two tag ranges are compressed into inline storage.
+ *
+ * This also covers a special case where both ranges contain
+ * `MTE_GRANULES_PER_PAGE / 2` tags and overflow the designated range size.
+ */
+static void test_two_ranges(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+	unsigned int i;
+	size_t r_len = 2;
+	unsigned char r_tags[2] = { 0xe, 0x0 };
+	unsigned short r_sizes[2];
+
+	for (i = 1; i < MTE_GRANULES_PER_PAGE; i++) {
+		r_sizes[0] = i;
+		r_sizes[1] = MTE_GRANULES_PER_PAGE - i;
+		mte_ranges_to_tags(r_tags, r_sizes, r_len, td->tags);
+		handle = compress_decompress_helper(test);
+		KUNIT_EXPECT_TRUE(test, mte_is_compressed(handle));
+	}
+}
+
+/*
+ * Test that a very small number of tag ranges ends up compressed into 8 bytes.
+ */
+static void test_compress_simple(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+
+	memset(td->tags, 0, MTE_PAGE_TAG_STORAGE);
+	td->tags[0] = 0xa0;
+	td->tags[1] = 0x0a;
+
+	handle = compress_decompress_helper(test);
+	/* Tags are stored inline. */
+	KUNIT_EXPECT_TRUE(test, mte_is_compressed(handle));
+}
+
+/*
+ * Test that a buffer containing @nranges ranges compresses into @exp_size
+ * bytes and decompresses into the original tag sequence.
+ */
+static void compress_range_helper(struct kunit *test, int nranges,
+				  bool exp_inl)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+
+	gen_tag_range_helper(td->tags, nranges);
+	handle = compress_decompress_helper(test);
+	KUNIT_EXPECT_EQ(test, mte_is_compressed(handle), exp_inl);
+}
+
+static inline size_t max_inline_ranges(void)
+{
+#if defined CONFIG_ARM64_4K_PAGES
+	return 6;
+#elif defined(CONFIG_ARM64_16K_PAGES)
+	return 5;
+#else
+	return 4;
+#endif
+}
+
+/*
+ * Test that every number of tag ranges is correctly compressed and
+ * decompressed.
+ */
+static void test_compress_ranges(struct kunit *test)
+{
+	unsigned int i;
+	bool exp_inl;
+
+	for (i = 1; i <= MTE_GRANULES_PER_PAGE; i++) {
+		exp_inl = i <= max_inline_ranges();
+		compress_range_helper(test, i, exp_inl);
+	}
+}
+
+/*
+ * Test that invalid handles are ignored by mte_decompress().
+ */
+static void test_decompress_invalid(struct kunit *test)
+{
+	void *handle1 = (void *)0xeb0b0b0100804020;
+	void *handle2 = (void *)0x6b0b0b010080402f;
+	struct test_data *td = test->priv;
+
+	/* handle1 has bit 0 set to 1. */
+	KUNIT_EXPECT_FALSE(test, mte_decompress(handle1, td->dtags));
+	/*
+	 * handle2 is an inline handle, but its largest_idx (bits 1..3)
+	 * is out of bounds for the inline storage.
+	 */
+	KUNIT_EXPECT_FALSE(test, mte_decompress(handle2, td->dtags));
+}
+
+/*
+ * Test that compressed inline tags cannot be confused with out-of-line
+ * pointers.
+ *
+ * Compressed values are written from bit 0 to bit 63, so the size of the last
+ * tag range initially ends up in the upper bits of the inline representation.
+ * Make sure mte_compress() rearranges the bits so that the resulting handle does
+ * not have 0b0111 as the upper four bits.
+ */
+static void test_upper_bits(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+	void *handle;
+	unsigned char r_tags[6] = { 7, 0, 7, 0, 7, 0 };
+	unsigned short r_sizes[6] = { 1, 1, 1, 1, 1, 1 };
+	size_t r_len;
+
+	/* Maximum number of ranges that can be encoded inline. */
+	r_len = max_inline_ranges();
+	/* Maximum range size possible, will be omitted. */
+	r_sizes[0] = MTE_GRANULES_PER_PAGE / 2 - 1;
+	/* A number close to r_sizes[0] that has most of its bits set. */
+	r_sizes[r_len - 1] = MTE_GRANULES_PER_PAGE - r_sizes[0] - r_len + 2;
+
+	mte_ranges_to_tags(r_tags, r_sizes, r_len, td->tags);
+	handle = compress_decompress_helper(test);
+	KUNIT_EXPECT_TRUE(test, mte_is_compressed(handle));
+}
+
+static void mtecomp_dealloc_testdata(struct test_data *td)
+{
+	kfree(td->tags);
+	kfree(td->dtags);
+	kfree(td->r_sizes);
+	kfree(td->r_tags);
+}
+
+static int mtecomp_test_init(struct kunit *test)
+{
+	struct test_data *td;
+
+	td = kmalloc(sizeof(struct test_data), GFP_KERNEL);
+	if (!td)
+		return 1;
+	td->tags = kmalloc(MTE_PAGE_TAG_STORAGE, GFP_KERNEL);
+	if (!td->tags)
+		goto error;
+	td->dtags = kmalloc(MTE_PAGE_TAG_STORAGE, GFP_KERNEL);
+	if (!td->dtags)
+		goto error;
+	td->r_len = MTE_GRANULES_PER_PAGE;
+	td->r_sizes = kmalloc_array(MTE_GRANULES_PER_PAGE,
+				    sizeof(unsigned short), GFP_KERNEL);
+	if (!td->r_sizes)
+		goto error;
+	td->r_tags = kmalloc(MTE_GRANULES_PER_PAGE, GFP_KERNEL);
+	if (!td->r_tags)
+		goto error;
+	test->priv = (void *)td;
+	return 0;
+error:
+	mtecomp_dealloc_testdata(td);
+	return 1;
+}
+
+static void mtecomp_test_exit(struct kunit *test)
+{
+	struct test_data *td = test->priv;
+
+	mtecomp_dealloc_testdata(td);
+}
+
+static struct kunit_case mtecomp_test_cases[] = {
+	KUNIT_CASE(test_tags_to_ranges_zero),
+	KUNIT_CASE(test_tags_to_ranges_simple),
+	KUNIT_CASE(test_tags_to_ranges_repeated),
+	KUNIT_CASE(test_tag_to_ranges_n),
+	KUNIT_CASE(test_compress_zero),
+	KUNIT_CASE(test_compress_nonzero),
+	KUNIT_CASE(test_two_ranges),
+	KUNIT_CASE(test_compress_simple),
+	KUNIT_CASE(test_compress_ranges),
+	KUNIT_CASE(test_decompress_invalid),
+	KUNIT_CASE(test_upper_bits),
+	{}
+};
+
+static struct kunit_suite mtecomp_test_suite = {
+	.name = "mtecomp",
+	.init = mtecomp_test_init,
+	.exit = mtecomp_test_exit,
+	.test_cases = mtecomp_test_cases,
+};
+kunit_test_suites(&mtecomp_test_suite);
+
+MODULE_IMPORT_NS(MTECOMP);
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Alexander Potapenko <glider@google.com>");
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 6/7] arm64: mte: add compression support to mteswap.c

[Alexander Potapenko]

Update mteswap.c to perform inline compression of memory tags when
possible.

If CONFIG_ARM64_MTE_COMP is enabled, mteswap.c will attempt to compress
saved tags for a struct page and store them directly in Xarray entry
instead of wasting heap space.

Soon after booting Android, tag compression saves ~2x memory previously
spent by mteswap.c on tag allocations. On a moderately loaded device with
~20% tagged pages, this leads to saving several megabytes of kernel heap:

  # cat /sys/kernel/debug/mteswap/stats
  8 bytes:	102496 allocations,	67302 deallocations
  128 bytes:	212234 allocations,	178278 deallocations
  uncompressed tag storage size:	8851200
  compressed tag storage size:	4346368

(statistics collection is introduced in the following patch)

Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>

---
 v10-mte:
  - added Catalin's Reviewed-by:

 v9:
  - as requested by Yury Norov, split off statistics collection into a
    separate patch
  - minor fixes

 v8:
  - adapt to the new compression API, abandon mteswap_{no,}comp.c
  - move stats collection to mteswap.c

 v5:
  - drop a dead variable from _mte_free_saved_tags() in mteswap_comp.c
  - ensure MTE compression works with arbitrary page sizes
  - update patch description

 v4:
  - minor code simplifications suggested by Andy Shevchenko, added
    missing header dependencies
  - changed compression API names to reflect modifications made to
    memcomp.h (as suggested by Yury Norov)

 v3:
  - Addressed comments by Andy Shevchenko in another patch:
   - fixed includes order
   - replaced u64 with unsigned long
   - added MODULE_IMPORT_NS(MTECOMP)
---
 arch/arm64/mm/mteswap.c | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git a/arch/arm64/mm/mteswap.c b/arch/arm64/mm/mteswap.c
index a31833e3ddc54..70f5c8ecd640d 100644
--- a/arch/arm64/mm/mteswap.c
+++ b/arch/arm64/mm/mteswap.c
@@ -6,6 +6,8 @@
 #include <linux/swap.h>
 #include <linux/swapops.h>
 #include <asm/mte.h>
+#include <asm/mtecomp.h>
+#include "mtecomp.h"
 
 static DEFINE_XARRAY(mte_pages);
 
@@ -17,12 +19,13 @@ void *mte_allocate_tag_storage(void)
 
 void mte_free_tag_storage(char *storage)
 {
-	kfree(storage);
+	if (!mte_is_compressed(storage))
+		kfree(storage);
 }
 
 int mte_save_tags(struct page *page)
 {
-	void *tag_storage, *ret;
+	void *tag_storage, *compressed_storage, *ret;
 
 	if (!page_mte_tagged(page))
 		return 0;
@@ -32,6 +35,11 @@ int mte_save_tags(struct page *page)
 		return -ENOMEM;
 
 	mte_save_page_tags(page_address(page), tag_storage);
+	compressed_storage = mte_compress(tag_storage);
+	if (compressed_storage) {
+		mte_free_tag_storage(tag_storage);
+		tag_storage = compressed_storage;
+	}
 
 	/* lookup the swap entry.val from the page */
 	ret = xa_store(&mte_pages, page_swap_entry(page).val, tag_storage,
@@ -50,13 +58,20 @@ int mte_save_tags(struct page *page)
 void mte_restore_tags(swp_entry_t entry, struct page *page)
 {
 	void *tags = xa_load(&mte_pages, entry.val);
+	void *tag_storage = NULL;
 
 	if (!tags)
 		return;
 
 	if (try_page_mte_tagging(page)) {
+		if (mte_is_compressed(tags)) {
+			tag_storage = mte_allocate_tag_storage();
+			mte_decompress(tags, tag_storage);
+			tags = tag_storage;
+		}
 		mte_restore_page_tags(page_address(page), tags);
 		set_page_mte_tagged(page);
+		mte_free_tag_storage(tag_storage);
 	}
 }
 
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

[PATCH v10-mte 7/7] arm64: mte: implement CONFIG_ARM64_MTE_SWAP_STATS

[Alexander Potapenko]

Provide a config to collect the usage statistics for ARM MTE tag
compression. This patch introduces allocation/deallocation counters
for buffers that were stored uncompressed (and thus occupy 128 bytes of
heap plus the Xarray overhead to store a pointer) and those that were
compressed into 8-byte pointers (effectively using 0 bytes of heap in
addition to the Xarray overhead).

The counters are exposed to the userspace via
/sys/kernel/debug/mteswap/stats:

  # cat /sys/kernel/debug/mteswap/stats
  8 bytes:      102496 allocations,     67302 deallocations
  128 bytes:    212234 allocations,     178278 deallocations
  uncompressed tag storage size:        8851200
  compressed tag storage size:  4346368

Suggested-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>

---

This patch was split off from the "arm64: mte: add compression support
to mteswap.c" patch
(https://lore.kernel.org/linux-arm-kernel/ZUVulBKVYK7cq2rJ@yury-ThinkPad/T/#m819ec30beb9de53d5c442f7e3247456f8966d88a)

v10-mte:
 - added Catalin's Acked-by:

v9:
 - add this patch, put the stats behind a separate config,
   mention /sys/kernel/debug/mteswap/stats in the documentation
---
 .../arch/arm64/mte-tag-compression.rst        | 12 +++
 arch/arm64/Kconfig                            | 15 +++
 arch/arm64/mm/mteswap.c                       | 93 ++++++++++++++++++-
 3 files changed, 118 insertions(+), 2 deletions(-)

diff --git a/Documentation/arch/arm64/mte-tag-compression.rst b/Documentation/arch/arm64/mte-tag-compression.rst
index 8fe6b51a9db6d..4c25b96f7d4b5 100644
--- a/Documentation/arch/arm64/mte-tag-compression.rst
+++ b/Documentation/arch/arm64/mte-tag-compression.rst
@@ -145,6 +145,18 @@ Tag compression and decompression implicitly rely on the fixed MTE tag size
 (4 bits) and number of tags per page. Should these values change, the algorithm
 may need to be revised.
 
+Stats
+=====
+
+When `CONFIG_ARM64_MTE_SWAP_STATS` is enabled, `arch/arm64/mm/mteswap.c` exports
+usage statistics for tag compression used when swapping tagged pages. The data
+can be accessed via debugfs::
+
+  # cat /sys/kernel/debug/mteswap/stats
+  8 bytes:        10438 allocations,      10417 deallocations
+  128 bytes:      26180 allocations,      26179 deallocations
+  uncompressed tag storage size:  2816
+  compressed tag storage size:    128
 
 Programming Interface
 =====================
diff --git a/arch/arm64/Kconfig b/arch/arm64/Kconfig
index 6a1397a96f2f0..49a786c7edadd 100644
--- a/arch/arm64/Kconfig
+++ b/arch/arm64/Kconfig
@@ -2100,6 +2100,21 @@ config ARM64_MTE_COMP_KUNIT_TEST
 	  be compressed into pointer-size values and correctly decompressed
 	  afterwards.
 
+config ARM64_MTE_SWAP_STATS
+	bool "Collect usage statistics of tag compression for swapped MTE tags"
+	default y
+	depends on ARM64_MTE && ARM64_MTE_COMP
+	help
+	  Collect usage statistics for ARM64 MTE tag compression during swapping.
+
+	  Adds allocation/deallocation counters for buffers that were stored
+	  uncompressed (and thus occupy 128 bytes of heap plus the Xarray
+	  overhead to store a pointer) and those that were compressed into
+	  8-byte pointers (effectively using 0 bytes of heap in addition to
+	  the Xarray overhead).
+	  The counters are exposed to the userspace via
+	  /sys/kernel/debug/mteswap/stats.
+
 config ARM64_SVE
 	bool "ARM Scalable Vector Extension support"
 	default y
diff --git a/arch/arm64/mm/mteswap.c b/arch/arm64/mm/mteswap.c
index 70f5c8ecd640d..1c6c78b9a9037 100644
--- a/arch/arm64/mm/mteswap.c
+++ b/arch/arm64/mm/mteswap.c
@@ -1,5 +1,6 @@
 // SPDX-License-Identifier: GPL-2.0-only
 
+#include <linux/debugfs.h>
 #include <linux/pagemap.h>
 #include <linux/xarray.h>
 #include <linux/slab.h>
@@ -11,16 +12,54 @@
 
 static DEFINE_XARRAY(mte_pages);
 
+enum mteswap_counters {
+	MTESWAP_CTR_INLINE = 0,
+	MTESWAP_CTR_OUTLINE,
+	MTESWAP_CTR_SIZE
+};
+
+#if defined(CONFIG_ARM64_MTE_SWAP_STATS)
+static atomic_long_t alloc_counters[MTESWAP_CTR_SIZE];
+static atomic_long_t dealloc_counters[MTESWAP_CTR_SIZE];
+
+static void inc_alloc_counter(int kind)
+{
+	atomic_long_inc(&alloc_counters[kind]);
+}
+
+static void inc_dealloc_counter(int kind)
+{
+	atomic_long_inc(&dealloc_counters[kind]);
+}
+#else
+static void inc_alloc_counter(int kind)
+{
+}
+
+static void inc_dealloc_counter(int kind)
+{
+}
+#endif
+
 void *mte_allocate_tag_storage(void)
 {
+	void *ret;
+
 	/* tags granule is 16 bytes, 2 tags stored per byte */
-	return kmalloc(MTE_PAGE_TAG_STORAGE, GFP_KERNEL);
+	ret = kmalloc(MTE_PAGE_TAG_STORAGE, GFP_KERNEL);
+	if (ret)
+		inc_alloc_counter(MTESWAP_CTR_OUTLINE);
+	return ret;
 }
 
 void mte_free_tag_storage(char *storage)
 {
-	if (!mte_is_compressed(storage))
+	if (!mte_is_compressed(storage)) {
 		kfree(storage);
+		inc_dealloc_counter(MTESWAP_CTR_OUTLINE);
+	} else {
+		inc_dealloc_counter(MTESWAP_CTR_INLINE);
+	}
 }
 
 int mte_save_tags(struct page *page)
@@ -39,6 +78,7 @@ int mte_save_tags(struct page *page)
 	if (compressed_storage) {
 		mte_free_tag_storage(tag_storage);
 		tag_storage = compressed_storage;
+		inc_alloc_counter(MTESWAP_CTR_INLINE);
 	}
 
 	/* lookup the swap entry.val from the page */
@@ -98,3 +138,52 @@ void mte_invalidate_tags_area(int type)
 	}
 	xa_unlock(&mte_pages);
 }
+
+#if defined(CONFIG_ARM64_MTE_SWAP_STATS)
+/* DebugFS interface. */
+static int stats_show(struct seq_file *seq, void *v)
+{
+	unsigned long total_mem_alloc = 0, total_mem_dealloc = 0;
+	unsigned long total_num_alloc = 0, total_num_dealloc = 0;
+	unsigned long sizes[2] = { 8, MTE_PAGE_TAG_STORAGE };
+	long alloc, dealloc;
+	unsigned long size;
+	int i;
+
+	for (i = 0; i < MTESWAP_CTR_SIZE; i++) {
+		alloc = atomic_long_read(&alloc_counters[i]);
+		dealloc = atomic_long_read(&dealloc_counters[i]);
+		total_num_alloc += alloc;
+		total_num_dealloc += dealloc;
+		size = sizes[i];
+		/*
+		 * Do not count 8-byte buffers towards compressed tag storage
+		 * size.
+		 */
+		if (i) {
+			total_mem_alloc += (size * alloc);
+			total_mem_dealloc += (size * dealloc);
+		}
+		seq_printf(seq,
+			   "%lu bytes:\t%lu allocations,\t%lu deallocations\n",
+			   size, alloc, dealloc);
+	}
+	seq_printf(seq, "uncompressed tag storage size:\t%lu\n",
+		   (total_num_alloc - total_num_dealloc) *
+			   MTE_PAGE_TAG_STORAGE);
+	seq_printf(seq, "compressed tag storage size:\t%lu\n",
+		   total_mem_alloc - total_mem_dealloc);
+	return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(stats);
+
+static int mteswap_init(void)
+{
+	struct dentry *mteswap_dir;
+
+	mteswap_dir = debugfs_create_dir("mteswap", NULL);
+	debugfs_create_file("stats", 0444, mteswap_dir, NULL, &stats_fops);
+	return 0;
+}
+module_init(mteswap_init);
+#endif
-- 
2.43.0.472.g3155946c3a-goog


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH v10-mte 7/7] arm64: mte: implement CONFIG_ARM64_MTE_SWAP_STATS

[Yury Norov]

On Thu, Dec 14, 2023 at 12:06:39PM +0100, Alexander Potapenko wrote:
> Provide a config to collect the usage statistics for ARM MTE tag
> compression. This patch introduces allocation/deallocation counters
> for buffers that were stored uncompressed (and thus occupy 128 bytes of
> heap plus the Xarray overhead to store a pointer) and those that were
> compressed into 8-byte pointers (effectively using 0 bytes of heap in
> addition to the Xarray overhead).
> 
> The counters are exposed to the userspace via
> /sys/kernel/debug/mteswap/stats:
> 
>   # cat /sys/kernel/debug/mteswap/stats
>   8 bytes:      102496 allocations,     67302 deallocations
>   128 bytes:    212234 allocations,     178278 deallocations
>   uncompressed tag storage size:        8851200
>   compressed tag storage size:  4346368
> 
> Suggested-by: Yury Norov <yury.norov@gmail.com>
> Signed-off-by: Alexander Potapenko <glider@google.com>
> Acked-by: Catalin Marinas <catalin.marinas@arm.com>

Reviewed-by: Yury Norov <yury.norov@gmail.com>

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel