From c3161248afdb42d1bf5e06a32041180cc4be457d Mon Sep 17 00:00:00 2001
From: Xavier Hernandez <jahernan@redhat.com>
Date: Fri, 6 Oct 2017 10:39:58 +0200
Subject: [PATCH 077/128] cluster/ec: add functions for stripe alignment
This patch removes old functions to align offsets and sizes
to stripe size boundaries and adds new ones to offer more
possibilities.
The new functions are:
* ec_adjust_offset_down()
Aligns a given offset to a multiple of the stripe size
equal or smaller than the initial one. It returns the
size of the gap between the aligned offset and the given
one.
* ec_adjust_offset_up()
Aligns a given offset to a multiple of the stripe size
equal or greater than the initial one. It returns the
size of the skipped region between the given offset and
the aligned one. If an overflow happens, the returned
valid has negative sign (but correct value) and the
offset is set to the maximum value (not aligned).
* ec_adjust_size_down()
Aligns the given size to a multiple of the stripe size
equal or smaller than the initial one. It returns the
size of the missed region between the aligned size and
the given one.
* ec_adjust_size_up()
Aligns the given size to a multiple of the stripe size
equal or greater than the initial one. It returns the
size of the gap between the given size and the aligned
one. If an overflow happens, the returned value has
negative sign (but correct value) and the size is set
to the maximum value (not aligned).
These functions have been defined in ec-helpers.h as static
inline since they are very small and compilers can optimize
them (specially the 'scale' argument).
upstream patch: https://review.gluster.org/#/c/18440/
>Change-Id: I4c91009ad02f76c73772034dfde27ee1c78a80d7
>Signed-off-by: Xavier Hernandez <jahernan@redhat.com>
BUG: 1499865
Change-Id: I4c91009ad02f76c73772034dfde27ee1c78a80d7
Signed-off-by: Sunil Kumar Acharya <sheggodu@redhat.com>
Reviewed-on: https://code.engineering.redhat.com/gerrit/123556
Tested-by: RHGS Build Bot <nigelb@redhat.com>
Reviewed-by: Javier Hernandez Juan <jahernan@redhat.com>
Reviewed-by: Atin Mukherjee <amukherj@redhat.com>
---
xlators/cluster/ec/src/ec-heal.c | 8 ++-
xlators/cluster/ec/src/ec-helpers.c | 29 ---------
xlators/cluster/ec/src/ec-helpers.h | 108 +++++++++++++++++++++++++++++++-
xlators/cluster/ec/src/ec-inode-read.c | 10 +--
xlators/cluster/ec/src/ec-inode-write.c | 13 ++--
xlators/cluster/ec/src/ec-locks.c | 8 +--
6 files changed, 129 insertions(+), 47 deletions(-)
diff --git a/xlators/cluster/ec/src/ec-heal.c b/xlators/cluster/ec/src/ec-heal.c
index a6de3ee..bc25015 100644
--- a/xlators/cluster/ec/src/ec-heal.c
+++ b/xlators/cluster/ec/src/ec-heal.c
@@ -1670,7 +1670,8 @@ ec_heal_data_find_direction (ec_t *ec, default_args_cbk_t *replies,
* well*/
if (check_ondisksize) {
- source_size = ec_adjust_size (ec, size[source], 1);
+ source_size = size[source];
+ ec_adjust_size_up (ec, &source_size, _gf_true);
for (i = 0; i < ec->nodes; i++) {
if (sources[i]) {
@@ -1983,7 +1984,7 @@ ec_rebuild_data (call_frame_t *frame, ec_t *ec, fd_t *fd, uint64_t size,
heal->fd = fd_ref (fd);
heal->xl = ec->xl;
heal->data = &barrier;
- size = ec_adjust_size (ec, size, 0);
+ ec_adjust_size_up (ec, &size, _gf_false);
heal->total_size = size;
heal->size = (128 * GF_UNIT_KB * (ec->self_heal_window_size));
/* We need to adjust the size to a multiple of the stripe size of the
@@ -2038,7 +2039,8 @@ __ec_heal_trim_sinks (call_frame_t *frame, ec_t *ec,
ret = 0;
goto out;
}
- trim_offset = ec_adjust_size (ec, size, 1);
+ trim_offset = size;
+ ec_adjust_offset_up (ec, &trim_offset, _gf_true);
ret = cluster_ftruncate (ec->xl_list, trim, ec->nodes, replies, output,
frame, ec->xl, fd, trim_offset, NULL);
for (i = 0; i < ec->nodes; i++) {
diff --git a/xlators/cluster/ec/src/ec-helpers.c b/xlators/cluster/ec/src/ec-helpers.c
index 64b010f..0c66948 100644
--- a/xlators/cluster/ec/src/ec-helpers.c
+++ b/xlators/cluster/ec/src/ec-helpers.c
@@ -799,35 +799,6 @@ ec_fd_t * ec_fd_get(fd_t * fd, xlator_t * xl)
return ctx;
}
-uint32_t ec_adjust_offset(ec_t * ec, off_t * offset, int32_t scale)
-{
- off_t head, tmp;
-
- tmp = *offset;
- head = tmp % ec->stripe_size;
- tmp -= head;
- if (scale)
- {
- tmp /= ec->fragments;
- }
-
- *offset = tmp;
-
- return head;
-}
-
-uint64_t ec_adjust_size(ec_t * ec, uint64_t size, int32_t scale)
-{
- size += ec->stripe_size - 1;
- size -= size % ec->stripe_size;
- if (scale)
- {
- size /= ec->fragments;
- }
-
- return size;
-}
-
gf_boolean_t
ec_is_internal_xattr (dict_t *dict, char *key, data_t *value, void *data)
{
diff --git a/xlators/cluster/ec/src/ec-helpers.h b/xlators/cluster/ec/src/ec-helpers.h
index 4d2145c..cfd7daa 100644
--- a/xlators/cluster/ec/src/ec-helpers.h
+++ b/xlators/cluster/ec/src/ec-helpers.h
@@ -55,8 +55,112 @@ ec_inode_t * ec_inode_get(inode_t * inode, xlator_t * xl);
ec_fd_t * __ec_fd_get(fd_t * fd, xlator_t * xl);
ec_fd_t * ec_fd_get(fd_t * fd, xlator_t * xl);
-uint32_t ec_adjust_offset(ec_t * ec, off_t * offset, int32_t scale);
-uint64_t ec_adjust_size(ec_t * ec, uint64_t size, int32_t scale);
+static inline uint32_t
+ec_adjust_size_down(ec_t *ec, uint64_t *value, gf_boolean_t scale)
+{
+ uint64_t head, tmp;
+
+ tmp = *value;
+ head = tmp % ec->stripe_size;
+ tmp -= head;
+
+ if (scale) {
+ tmp /= ec->fragments;
+ }
+
+ *value = tmp;
+
+ return (uint32_t)head;
+}
+
+/* This function can cause an overflow if the passed value is too near to the
+ * uint64_t limit. If this happens, it returns the tail in negative form and
+ * the value is set to UINT64_MAX. */
+static inline int32_t
+ec_adjust_size_up(ec_t *ec, uint64_t *value, gf_boolean_t scale)
+{
+ uint64_t tmp;
+ int32_t tail;
+
+ tmp = *value;
+ /* We first adjust the value down. This never causes overflow. */
+ tail = ec_adjust_size_down(ec, &tmp, scale);
+
+ /* If the value was already aligned, tail will be 0 and nothing else
+ * needs to be done. */
+ if (tail != 0) {
+ /* Otherwise, we need to compute the real tail and adjust the
+ * returned value to the next stripe. */
+ tail = ec->stripe_size - tail;
+ if (scale) {
+ tmp += ec->fragment_size;
+ } else {
+ tmp += ec->stripe_size;
+ /* If no scaling is requested there's a posibility of
+ * overflow. */
+ if (tmp < ec->stripe_size) {
+ tmp = UINT64_MAX;
+ tail = -tail;
+ }
+ }
+ }
+
+ *value = tmp;
+
+ return tail;
+}
+
+/* This function is equivalent to ec_adjust_size_down() but with a potentially
+ * different parameter size (off_t vs uint64_t). */
+static inline uint32_t
+ec_adjust_offset_down(ec_t *ec, off_t *value, gf_boolean_t scale)
+{
+ off_t head, tmp;
+
+ tmp = *value;
+ head = tmp % ec->stripe_size;
+ tmp -= head;
+
+ if (scale) {
+ tmp /= ec->fragments;
+ }
+
+ *value = tmp;
+
+ return (uint32_t)head;
+}
+
+/* This function is equivalent to ec_adjust_size_up() but with a potentially
+ * different parameter size (off_t vs uint64_t). */
+static inline int32_t
+ec_adjust_offset_up(ec_t *ec, off_t *value, gf_boolean_t scale)
+{
+ uint64_t tail, tmp;
+
+ /* An offset is a signed type that can only have positive values, so
+ * we take advantage of this to avoid overflows. We simply convert it
+ * to an unsigned integer and operate normally. This won't cause an
+ * overflow. Overflow is only checked when converting back to an
+ * off_t. */
+ tmp = *value;
+ tail = ec->stripe_size;
+ tail -= (tmp + tail - 1) % tail + 1;
+ tmp += tail;
+ if (scale) {
+ /* If we are scaling, we'll never get an overflow. */
+ tmp /= ec->fragments;
+ } else {
+ /* Check if there has been an overflow. */
+ if ((off_t)tmp < 0) {
+ tmp = (1ULL << (sizeof(off_t) * 8 - 1)) - 1ULL;
+ tail = -tail;
+ }
+ }
+
+ *value = (off_t)tmp;
+
+ return (int32_t)tail;
+}
static inline int32_t ec_is_power_of_2(uint32_t value)
{
diff --git a/xlators/cluster/ec/src/ec-inode-read.c b/xlators/cluster/ec/src/ec-inode-read.c
index d925e82..829f47f 100644
--- a/xlators/cluster/ec/src/ec-inode-read.c
+++ b/xlators/cluster/ec/src/ec-inode-read.c
@@ -1356,9 +1356,10 @@ int32_t ec_manager_readv(ec_fop_data_t * fop, int32_t state)
{
case EC_STATE_INIT:
fop->user_size = fop->size;
- fop->head = ec_adjust_offset(fop->xl->private, &fop->offset, 1);
- fop->size = ec_adjust_size(fop->xl->private, fop->size + fop->head,
- 1);
+ fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
+ _gf_true);
+ fop->size += fop->head;
+ ec_adjust_size_up(fop->xl->private, &fop->size, _gf_true);
/* Fall through */
@@ -1561,7 +1562,8 @@ int32_t ec_manager_seek(ec_fop_data_t *fop, int32_t state)
switch (state) {
case EC_STATE_INIT:
fop->user_size = fop->offset;
- fop->head = ec_adjust_offset(fop->xl->private, &fop->offset, 1);
+ fop->head = ec_adjust_offset_down(fop->xl->private, &fop->offset,
+ _gf_true);
/* Fall through */
diff --git a/xlators/cluster/ec/src/ec-inode-write.c b/xlators/cluster/ec/src/ec-inode-write.c
index 68bea1a..3ed9b2a 100644
--- a/xlators/cluster/ec/src/ec-inode-write.c
+++ b/xlators/cluster/ec/src/ec-inode-write.c
@@ -870,8 +870,10 @@ int32_t ec_manager_fallocate(ec_fop_data_t *fop, int32_t state)
return EC_STATE_REPORT;
}
fop->user_size = fop->offset + fop->size;
- fop->head = ec_adjust_offset (fop->xl->private, &fop->offset, 1);
- fop->size = ec_adjust_size (fop->xl->private, fop->head + fop->size, 1);
+ fop->head = ec_adjust_offset_down (fop->xl->private, &fop->offset,
+ _gf_true);
+ fop->size += fop->head;
+ ec_adjust_size_up (fop->xl->private, &fop->size, _gf_true);
/* Fall through */
@@ -1145,7 +1147,7 @@ int32_t ec_manager_truncate(ec_fop_data_t * fop, int32_t state)
{
case EC_STATE_INIT:
fop->user_size = fop->offset;
- fop->offset = ec_adjust_size(fop->xl->private, fop->offset, 1);
+ ec_adjust_offset_up(fop->xl->private, &fop->offset, _gf_true);
/* Fall through */
@@ -1508,8 +1510,9 @@ ec_writev_prepare_buffers(ec_t *ec, ec_fop_data_t *fop)
int32_t err;
fop->user_size = iov_length(fop->vector, fop->int32);
- fop->head = ec_adjust_offset(ec, &fop->offset, 0);
- fop->size = ec_adjust_size(ec, fop->user_size + fop->head, 0);
+ fop->head = ec_adjust_offset_down(ec, &fop->offset, _gf_false);
+ fop->size = fop->user_size + fop->head;
+ ec_adjust_size_up(ec, &fop->size, _gf_false);
if ((fop->int32 != 1) || (fop->head != 0) ||
(fop->size > fop->user_size) ||
diff --git a/xlators/cluster/ec/src/ec-locks.c b/xlators/cluster/ec/src/ec-locks.c
index ff09852..996035d 100644
--- a/xlators/cluster/ec/src/ec-locks.c
+++ b/xlators/cluster/ec/src/ec-locks.c
@@ -572,10 +572,10 @@ int32_t ec_manager_inodelk(ec_fop_data_t * fop, int32_t state)
switch (state)
{
case EC_STATE_INIT:
- fop->flock.l_len += ec_adjust_offset(fop->xl->private,
- &fop->flock.l_start, 1);
- fop->flock.l_len = ec_adjust_size(fop->xl->private,
- fop->flock.l_len, 1);
+ fop->flock.l_len += ec_adjust_offset_down(fop->xl->private,
+ &fop->flock.l_start,
+ _gf_true);
+ ec_adjust_offset_up(fop->xl->private, &fop->flock.l_len, _gf_true);
if ((fop->int32 == F_SETLKW) && (fop->flock.l_type != F_UNLCK))
{
fop->uint32 = EC_LOCK_MODE_ALL;
--
1.8.3.1