f3/libprobe.c
Michel Machado 902db294ef f3probe: align buffers with devices' block sizes
This patch should address issue discussed here:
https://github.com/AltraMayor/f3/issues/13

Reported-by: Dingyuan Wang
2015-02-14 10:36:07 -05:00

398 lines
11 KiB
C

#include <stdlib.h>
#include <string.h>
#include <stdbool.h>
#include <assert.h>
#include <math.h>
#include <errno.h>
#include "libutils.h"
#include "libprobe.h"
static inline int equal_blk(struct device *dev, const char *b1, const char *b2)
{
return !memcmp(b1, b2, dev_get_block_size(dev));
}
/* Return true if @b1 and b2 are at most @tolerance_byte bytes different. */
static int similar_blk(struct device *dev, const char *b1, const char *b2,
int tolerance_byte)
{
const int block_size = dev_get_block_size(dev);
int i;
for (i = 0; i < block_size; i++) {
if (*b1 != *b2) {
tolerance_byte--;
if (tolerance_byte <= 0)
return false;
}
b1++;
b2++;
}
return true;
}
/* Return true if the block at @pos is damaged. */
static int test_block(struct device *dev,
const char *stamp_blk, char *probe_blk, uint64_t pos)
{
/* Write block. */
if (dev_write_block(dev, stamp_blk, pos) &&
dev_write_block(dev, stamp_blk, pos))
return true;
/* Reset. */
if (dev_reset(dev) && dev_reset(dev))
return true;
/*
* Test block.
*/
if (dev_read_block(dev, probe_blk, pos) &&
dev_read_block(dev, probe_blk, pos))
return true;
if (equal_blk(dev, stamp_blk, probe_blk))
return false;
/* Save time with certainly damaged blocks. */
if (!similar_blk(dev, stamp_blk, probe_blk, 8)) {
/* The probe block is damaged. */
return true;
}
/* The probe block seems to be damaged.
* Trying a second time...
*/
return dev_write_and_reset(dev, stamp_blk, pos) ||
dev_read_block(dev, probe_blk, pos) ||
!equal_blk(dev, stamp_blk, probe_blk);
}
/* Minimum size of the memory chunk used to build flash drives.
* It must be a power of two.
*/
static inline uint64_t initial_high_bit_block(struct device *dev)
{
int block_order = dev_get_block_order(dev);
assert(block_order <= 20);
return 1ULL << (20 - block_order);
}
/* Caller must guarantee that the left bock is good, and written. */
static int search_wrap(struct device *dev,
uint64_t left_pos, uint64_t *pright_pos,
const char *stamp_blk, char *probe_blk)
{
uint64_t high_bit = initial_high_bit_block(dev);
uint64_t pos = high_bit + left_pos;
/* The left block must be in the first memory chunk. */
assert(left_pos < high_bit);
/* Check that the drive has at least one memory chunk. */
assert((high_bit - 1) <= *pright_pos);
while (pos < *pright_pos) {
if (dev_read_block(dev, probe_blk, pos) &&
dev_read_block(dev, probe_blk, pos))
return true;
/* XXX Deal with flipped bit on reception. */
if (equal_blk(dev, stamp_blk, probe_blk)) {
/* XXX Test wraparound hypothesis. */
*pright_pos = high_bit - 1;
return false;
}
high_bit <<= 1;
pos = high_bit + left_pos;
}
return false;
}
#define MAX_N_BLOCK_ORDER 10
static uint64_t estimate_best_n_block(struct device *dev)
{
uint64_t write_count, write_time_us;
uint64_t reset_count, reset_time_us;
double t_w_us, t_2w_us, t_r_us;
uint64_t n_block_order;
perf_device_sample(dev, NULL, NULL, &write_count, &write_time_us,
&reset_count, &reset_time_us);
if (write_count < 3 || reset_count < 2) {
/* There is not enough measurements. */
return (1 << 2) - 1;
}
/* Let 2^n be the total number of blocks on the drive.
* Let p be the total number of passes.
* Let w = (2^m - 1) be the number of blocks written on each pass,
* where m >= 1.
*
* A pass is an iteration of the loop in search_edge(), that is,
* a call to write_test_blocks(), dev_reset(), and probe_test_blocks().
*
* The reason to have w = (2^m - 1) instead of w = 2^m is because
* the former leads to a clean relationship between n, p, and m
* when m is constant: 2^n / (w + 1)^p = 1 => p = n/m
*
* Let Tr be the time to reset the device.
* Let Tw be the time to write a block to @dev.
* Let Tw' be the time to write a block to the underlying device
* of @dev, that is, without overhead due to chaining multiple
* struct device. For example, when struct safe_device is used
* Tw > Tw'.
* Let Trd be the time to read a block from @dev.
*
* Notice that each single-block pass reduces the search space in half,
* and that to reduce the search space in half writing blocks,
* one has to increase m of one.
*
* Thus, in order to be better writing more blocks than
* going for another pass, the following relation must be true:
*
* Tr + Tw + Tw' >= (w - 1)(Tw + Tw')
*
* The relation above assumes Trd = 0.
*
* The left side of the relation above is the time to do _another_
* pass writing a single block, whereas the right side is the time to
* stay in the same pass and write (w - 1) more blocks.
* In order words, if there is no advantage to write more blocks,
* we stick to single-block passes.
*
* Tw' is there to account for any operation that writes
* the blocks back (e.g. using struct safe_device), otherwise
* processing operations related per written blocks that is not
* being accounted for (e.g. reading the blocks back to test).
*
* Solving the relation for w: w <= Tr/(Tw + Tw') + 2
*
* However, we are not interested in any w, but only those of
* of the form (2^m - 1) to make sure that we are not better off
* calling another pass. Thus, solving the previous relation for m:
*
* m <= log_2(Tr/(Tw + Tw') + 3)
*
* We approximate Tw' making it equal to Tw.
*/
t_w_us = (double)write_time_us / write_count;
t_r_us = (double)reset_time_us / reset_count;
t_2w_us = t_w_us > 0. ? 2. * t_w_us : 1.; /* Avoid zero division. */
n_block_order = ilog2(round(t_r_us / t_2w_us + 3.));
/* Bound the maximum number of blocks per pass to limit
* the necessary amount of memory struct safe_device pre-allocates.
*/
if (n_block_order > MAX_N_BLOCK_ORDER)
n_block_order = MAX_N_BLOCK_ORDER;
return (1 << n_block_order) - 1;
}
/* Write blocks whose offsets are after @left_pos but
* less or equal to @right_pos.
*/
static int write_test_blocks(struct device *dev, const char *stamp_blk,
uint64_t left_pos, uint64_t right_pos,
uint64_t *pa, uint64_t *pb, uint64_t *pmax_idx)
{
uint64_t pos, last_pos;
uint64_t n_block = estimate_best_n_block(dev);
assert(n_block >= 1);
/* Find coeficients of function a*idx + b where idx <= max_idx. */
assert(left_pos < right_pos);
*pb = left_pos + 1;
*pa = round((right_pos - *pb) / (n_block + 1.));
*pa = !*pa ? 1ULL : *pa;
*pmax_idx = (right_pos - *pb) / *pa;
if (*pmax_idx >= n_block) {
/* Shift the zero of the function to the right.
* This avoids picking the leftmost block when a more
* informative block to the right is available.
* This also biases toward righter blocks,
* what improves the time to test good flash drives.
*/
*pb += *pa;
*pmax_idx = n_block - 1;
}
last_pos = *pa * *pmax_idx + *pb;
assert(last_pos <= right_pos);
/* Write test blocks. */
for (pos = *pb; pos <= last_pos; pos += *pa)
if (dev_write_block(dev, stamp_blk, pos) &&
dev_write_block(dev, stamp_blk, pos))
return true;
return false;
}
/* Return true if the test block at @pos is damaged. */
static int test_test_block(struct device *dev,
const char *stamp_blk, char *probe_blk, uint64_t pos)
{
if (dev_read_block(dev, probe_blk, pos) &&
dev_read_block(dev, probe_blk, pos))
return true;
return !equal_blk(dev, stamp_blk, probe_blk);
}
static int probe_test_blocks(struct device *dev,
const char *stamp_blk, char *probe_blk,
uint64_t *pleft_pos, uint64_t *pright_pos,
uint64_t a, uint64_t b, uint64_t max_idx)
{
/* Signed variables. */
int64_t left_idx = 0;
int64_t right_idx = max_idx;
int64_t idx = right_idx;
while (left_idx <= right_idx) {
uint64_t pos = a * idx + b;
if (test_test_block(dev, stamp_blk, probe_blk, pos)) {
right_idx = idx - 1;
*pright_pos = pos;
} else {
left_idx = idx + 1;
*pleft_pos = pos;
}
idx = (left_idx + right_idx) / 2;
}
return false;
}
/* Caller must guarantee that the left bock is good, and written. */
static int search_edge(struct device *dev,
uint64_t *pleft_pos, uint64_t right_pos,
const char *stamp_blk, char *probe_blk)
{
uint64_t gap = right_pos - *pleft_pos;
uint64_t prv_gap = gap + 1;
while (prv_gap > gap && gap >= 1) {
uint64_t a, b, max_idx;
if (write_test_blocks(dev, stamp_blk, *pleft_pos, right_pos,
&a, &b, &max_idx))
return true;
/* Reset. */
if (dev_reset(dev) && dev_reset(dev))
return true;
if (probe_test_blocks(dev, stamp_blk, probe_blk,
pleft_pos, &right_pos, a, b, max_idx))
return true;
prv_gap = gap;
gap = right_pos - *pleft_pos;
}
return false;
}
/* XXX Write random data to make it harder for fake chips to become "smarter".
* There would be a random seed.
* Buffer cannot be all 0x00 or all 0xFF.
*/
static void fill_buffer(char *buf, int len)
{
memset(buf, 0xAA, len);
}
int probe_device_max_blocks(struct device *dev)
{
uint64_t num_blocks = dev_get_size_byte(dev) >>
dev_get_block_order(dev);
int n = ceiling_log2(num_blocks);
/* Make sure that there is no overflow in the formula below.
* The number 10 is arbitrary here, that is, it's not tight.
*/
assert(MAX_N_BLOCK_ORDER < sizeof(int) - 10);
return
/* search_wrap() */
1 +
/* Search_edge()
*
* The number of used blocks is (p * w); see comments in
* estimate_best_n_block() for the definition of the variables.
*
* p * w = n/m * (2^m - 1) < n/m * 2^m = n * (2^m / m)
*
* Let f(m) be 2^m / m. One can prove that f(m + 1) >= f(m)
* for all m >= 1. Therefore, the following bound is true.
*
* p * w < n * f(max_m)
*/
((n << MAX_N_BLOCK_ORDER) / MAX_N_BLOCK_ORDER);
}
/* XXX Properly handle read and write errors.
* Review each assert to check if them can be removed.
*/
int probe_device(struct device *dev, uint64_t *preal_size_byte,
uint64_t *pannounced_size_byte, int *pwrap, int *pblock_order)
{
uint64_t dev_size_byte = dev_get_size_byte(dev);
const int block_size = dev_get_block_size(dev);
const int block_order = dev_get_block_order(dev);
char stack[align_head(block_order) + (2 << block_order)];
char *stamp_blk, *probe_blk;
/* XXX Don't write at the very beginning of the card to avoid
* losing the partition table.
* But write at a random locations to make harder for fake chips
* to become "smarter".
* And try a couple of blocks if they keep failing.
*/
uint64_t left_pos = 10;
uint64_t right_pos = (dev_size_byte >> block_order) - 1;
struct device *pdev;
assert(dev_size_byte % block_size == 0);
assert(left_pos < right_pos);
pdev = create_perf_device(dev);
if (!pdev)
return -ENOMEM;
/* Aligning these pointers is necessary to directly read and write
* the block device.
* For the file device, this is superfluous.
*/
stamp_blk = align_mem(stack, block_order);
probe_blk = stamp_blk + block_size;
fill_buffer(stamp_blk, block_size);
/* Make sure that there is at least a good block at the beginning
* of the drive.
*/
if (test_block(pdev, stamp_blk, probe_blk, left_pos))
goto bad;
if (search_wrap(pdev, left_pos, &right_pos, stamp_blk, probe_blk))
goto bad;
if (search_edge(pdev, &left_pos, right_pos, stamp_blk, probe_blk))
goto bad;
*preal_size_byte = (left_pos + 1) << block_order;
*pannounced_size_byte = dev_size_byte;
*pwrap = ceiling_log2(right_pos << block_order);
*pblock_order = block_order;
goto out;
bad:
*preal_size_byte = 0;
*pannounced_size_byte = dev_size_byte;
*pwrap = ceiling_log2(dev_size_byte);
*pblock_order = block_order;
out:
pdev_detach_and_free(pdev);
return 0;
}