Skip to content

L
linux
Commit 95a0f10c authored by Lars Ellenberg's avatar Lars Ellenberg Committed by Philipp Reisner
Browse files
Options

drbd: store in-core bitmap little endian, regardless of architecture 

Our on-disk bitmap is a little endian bitstream.
Up to now, we have stored the in-core copy of that in
native endian, applying byte order conversion when necessary.

Instead, keep the bitmap pages little endian, as they are read from disk,
and use the generic_*_le_bit family of functions.
Signed-off-by: default avatarPhilipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: default avatarLars Ellenberg <lars.ellenberg@linbit.com>
parent 7777a8ba
Hide whitespace changes
Inline Side-by-side
Showing with 84 additions and 82 deletions
drivers/block/drbd/drbd_bitmap.c
View file @ 95a0f10c
......@@ -30,6 +30,7 @@
#include <asm/kmap_types.h>
#include "drbd_int.h"
/* OPAQUE outside this file!
* interface defined in drbd_int.h
......@@ -154,6 +155,14 @@ void drbd_bm_unlock(struct drbd_conf *mdev)
mutex_unlock(&b->bm_change);
}
static unsigned int bm_bit_to_page_idx(struct drbd_bitmap *b, u64 bitnr)
{
/* page_nr = (bitnr/8) >> PAGE_SHIFT; */
unsigned int page_nr = bitnr >> (PAGE_SHIFT + 3);
BUG_ON(page_nr >= b->bm_number_of_pages);
return page_nr;
}
/* word offset to long pointer */
static unsigned long *__bm_map_paddr(struct drbd_bitmap *b, unsigned long offset, const enum km_type km)
{
......@@ -168,6 +177,17 @@ static unsigned long *__bm_map_paddr(struct drbd_bitmap *b, unsigned long offset
return (unsigned long *) kmap_atomic(page, km);
}
static unsigned long *__bm_map_pidx(struct drbd_bitmap *b, unsigned int idx, const enum km_type km)
{
struct page *page = b->bm_pages[idx];
return (unsigned long *) kmap_atomic(page, km);
}
static unsigned long *bm_map_pidx(struct drbd_bitmap *b, unsigned int idx)
{
return __bm_map_pidx(b, idx, KM_IRQ1);
}
static unsigned long * bm_map_paddr(struct drbd_bitmap *b, unsigned long offset)
{
return __bm_map_paddr(b, offset, KM_IRQ1);
......@@ -329,22 +349,42 @@ void drbd_bm_cleanup(struct drbd_conf *mdev)
* this masks out the remaining bits.
* Returns the number of bits cleared.
*/
#define BITS_PER_PAGE (1UL << (PAGE_SHIFT + 3))
#define BITS_PER_PAGE_MASK (BITS_PER_PAGE - 1)
#define BITS_PER_LONG_MASK (BITS_PER_LONG - 1)
static int bm_clear_surplus(struct drbd_bitmap *b)
{
const unsigned long mask = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) - 1;
size_t w = b->bm_bits >> LN2_BPL;
int cleared = 0;
unsigned long mask;
unsigned long *p_addr, *bm;
int tmp;
int cleared = 0;
p_addr = bm_map_paddr(b, w);
bm = p_addr + MLPP(w);
if (w < b->bm_words) {
/* number of bits modulo bits per page */
tmp = (b->bm_bits & BITS_PER_PAGE_MASK);
/* mask the used bits of the word containing the last bit */
mask = (1UL << (tmp & BITS_PER_LONG_MASK)) -1;
/* bitmap is always stored little endian,
* on disk and in core memory alike */
mask = cpu_to_lel(mask);
/* because of the "extra long to catch oob access" we allocate in
* drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page
* containing the last _relevant_ bitmap word */
p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1));
bm = p_addr + (tmp/BITS_PER_LONG);
if (mask) {
/* If mask != 0, we are not exactly aligned, so bm now points
* to the long containing the last bit.
* If mask == 0, bm already points to the word immediately
* after the last (long word aligned) bit. */
cleared = hweight_long(*bm & ~mask);
*bm &= mask;
w++; bm++;
bm++;
}
if (w < b->bm_words) {
if (BITS_PER_LONG == 32 && ((bm - p_addr) & 1) == 1) {
/* on a 32bit arch, we may need to zero out
* a padding long to align with a 64bit remote */
cleared += hweight_long(*bm);
*bm = 0;
}
......@@ -354,24 +394,41 @@ static int bm_clear_surplus(struct drbd_bitmap *b)
static void bm_set_surplus(struct drbd_bitmap *b)
{
const unsigned long mask = (1UL << (b->bm_bits & (BITS_PER_LONG-1))) - 1;
size_t w = b->bm_bits >> LN2_BPL;
unsigned long mask;
unsigned long *p_addr, *bm;
p_addr = bm_map_paddr(b, w);
bm = p_addr + MLPP(w);
if (w < b->bm_words) {
int tmp;
/* number of bits modulo bits per page */
tmp = (b->bm_bits & BITS_PER_PAGE_MASK);
/* mask the used bits of the word containing the last bit */
mask = (1UL << (tmp & BITS_PER_LONG_MASK)) -1;
/* bitmap is always stored little endian,
* on disk and in core memory alike */
mask = cpu_to_lel(mask);
/* because of the "extra long to catch oob access" we allocate in
* drbd_bm_resize, bm_number_of_pages -1 is not necessarily the page
* containing the last _relevant_ bitmap word */
p_addr = bm_map_pidx(b, bm_bit_to_page_idx(b, b->bm_bits - 1));
bm = p_addr + (tmp/BITS_PER_LONG);
if (mask) {
/* If mask != 0, we are not exactly aligned, so bm now points
* to the long containing the last bit.
* If mask == 0, bm already points to the word immediately
* after the last (long word aligned) bit. */
*bm |= ~mask;
bm++; w++;
bm++;
}
if (w < b->bm_words) {
*bm = ~(0UL);
if (BITS_PER_LONG == 32 && ((bm - p_addr) & 1) == 1) {
/* on a 32bit arch, we may need to zero out
* a padding long to align with a 64bit remote */
*bm = ~0UL;
}
bm_unmap(p_addr);
}
static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endian)
static unsigned long bm_count_bits(struct drbd_bitmap *b)
{
unsigned long *p_addr, *bm, offset = 0;
unsigned long bits = 0;
......@@ -389,10 +446,6 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia
p_addr = __bm_map_paddr(b, offset, KM_USER0);
bm = p_addr + MLPP(offset);
while (i--) {
#ifndef __LITTLE_ENDIAN
if (swap_endian)
*bm = lel_to_cpu(*bm);
#endif
bits += hweight_long(*bm++);
}
offset += do_now;
......@@ -415,16 +468,6 @@ static unsigned long __bm_count_bits(struct drbd_bitmap *b, const int swap_endia
return bits;
}
static unsigned long bm_count_bits(struct drbd_bitmap *b)
{
return __bm_count_bits(b, 0);
}
static unsigned long bm_count_bits_swap_endian(struct drbd_bitmap *b)
{
return __bm_count_bits(b, 1);
}
/* offset and len in long words.*/
static void bm_memset(struct drbd_bitmap *b, size_t offset, int c, size_t len)
{
......@@ -662,7 +705,7 @@ void drbd_bm_merge_lel(struct drbd_conf *mdev, size_t offset, size_t number,
offset += do_now;
while (do_now--) {
bits = hweight_long(*bm);
word = *bm | lel_to_cpu(*buffer++);
word = *bm | *buffer++;
*bm++ = word;
b->bm_set += hweight_long(word) - bits;
}
......@@ -709,7 +752,7 @@ void drbd_bm_get_lel(struct drbd_conf *mdev, size_t offset, size_t number,
bm = p_addr + MLPP(offset);
offset += do_now;
while (do_now--)
*buffer++ = cpu_to_lel(*bm++);
*buffer++ = *bm++;
bm_unmap(p_addr);
}
}
......@@ -795,39 +838,6 @@ static void bm_page_io_async(struct drbd_conf *mdev, struct drbd_bitmap *b, int
}
}
# if defined(__LITTLE_ENDIAN)
/* nothing to do, on disk == in memory */
# define bm_cpu_to_lel(x) ((void)0)
# else
static void bm_cpu_to_lel(struct drbd_bitmap *b)
{
/* need to cpu_to_lel all the pages ...
* this may be optimized by using
* cpu_to_lel(-1) == -1 and cpu_to_lel(0) == 0;
* the following is still not optimal, but better than nothing */
unsigned int i;
unsigned long *p_addr, *bm;
if (b->bm_set == 0) {
/* no page at all; avoid swap if all is 0 */
i = b->bm_number_of_pages;
} else if (b->bm_set == b->bm_bits) {
/* only the last page */
i = b->bm_number_of_pages - 1;
} else {
/* all pages */
i = 0;
}
for (; i < b->bm_number_of_pages; i++) {
p_addr = kmap_atomic(b->bm_pages[i], KM_USER0);
for (bm = p_addr; bm < p_addr + PAGE_SIZE/sizeof(long); bm++)
*bm = cpu_to_lel(*bm);
kunmap_atomic(p_addr, KM_USER0);
}
}
# endif
/* lel_to_cpu == cpu_to_lel */
# define bm_lel_to_cpu(x) bm_cpu_to_lel(x)
/*
* bm_rw: read/write the whole bitmap from/to its on disk location.
*/
......@@ -847,10 +857,6 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local)
bm_words = drbd_bm_words(mdev);
num_pages = (bm_words*sizeof(long) + PAGE_SIZE-1) >> PAGE_SHIFT;
/* on disk bitmap is little endian */
if (rw == WRITE)
bm_cpu_to_lel(b);
now = jiffies;
atomic_set(&b->bm_async_io, num_pages);
__clear_bit(BM_MD_IO_ERROR, &b->bm_flags);
......@@ -869,13 +875,9 @@ static int bm_rw(struct drbd_conf *mdev, int rw) __must_hold(local)
now = jiffies;
if (rw == WRITE) {
/* swap back endianness */
bm_lel_to_cpu(b);
/* flush bitmap to stable storage */
drbd_md_flush(mdev);
} else /* rw == READ */ {
/* just read, if necessary adjust endianness */
b->bm_set = bm_count_bits_swap_endian(b);
b->bm_set = bm_count_bits(b);
dev_info(DEV, "recounting of set bits took additional %lu jiffies\n",
jiffies - now);
}
......@@ -969,9 +971,9 @@ static unsigned long __bm_find_next(struct drbd_conf *mdev, unsigned long bm_fo,
p_addr = __bm_map_paddr(b, offset, km);
if (find_zero_bit)
i = find_next_zero_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK);
i = generic_find_next_zero_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK);
else
i = find_next_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK);
i = generic_find_next_le_bit(p_addr, PAGE_SIZE*8, bm_fo & BPP_MASK);
__bm_unmap(p_addr, km);
if (i < PAGE_SIZE*8) {
......@@ -1064,9 +1066,9 @@ static int __bm_change_bits_to(struct drbd_conf *mdev, const unsigned long s,
last_page_nr = page_nr;
}
if (val)
c += (0 == __test_and_set_bit(bitnr & BPP_MASK, p_addr));
c += (0 == generic___test_and_set_le_bit(bitnr & BPP_MASK, p_addr));
else
c -= (0 != __test_and_clear_bit(bitnr & BPP_MASK, p_addr));
c -= (0 != generic___test_and_clear_le_bit(bitnr & BPP_MASK, p_addr));
}
if (p_addr)
__bm_unmap(p_addr, km);
......@@ -1211,7 +1213,7 @@ int drbd_bm_test_bit(struct drbd_conf *mdev, const unsigned long bitnr)
if (bitnr < b->bm_bits) {
unsigned long offset = bitnr>>LN2_BPL;
p_addr = bm_map_paddr(b, offset);
i = test_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0;
i = generic_test_le_bit(bitnr & BPP_MASK, p_addr) ? 1 : 0;
bm_unmap(p_addr);
} else if (bitnr == b->bm_bits) {
i = -1;
......@@ -1255,7 +1257,7 @@ int drbd_bm_count_bits(struct drbd_conf *mdev, const unsigned long s, const unsi
ERR_IF (bitnr >= b->bm_bits) {
dev_err(DEV, "bitnr=%lu bm_bits=%lu\n", bitnr, b->bm_bits);
} else {
c += (0 != test_bit(bitnr - (page_nr << (PAGE_SHIFT+3)), p_addr));
c += (0 != generic_test_le_bit(bitnr - (page_nr << (PAGE_SHIFT+3)), p_addr));
}
}
if (p_addr)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment
GitLab Nexedi Edition | About GitLab | About Nexedi | 沪ICP备2021021310号-2 | 沪ICP备2021021310号-7