Resend v5 by a new mail address(huaweicloud.com) because old
address(huawei.com)has some problem that emails can end up in spam.
Please let me know if anyone still see this patchset end up in spam.
Changes in v6:
- rename parameter in patch 3
- add comments and reviewed tag for patch 4
Changes in v5:
- add comments in patch 4
- clear bytes/io_skipped in throtl_start_new_slice_with_credit() in
patch 4
- and cleanup patches 5-8
Changes in v4:
- add reviewed-by tag for patch 1
- add patch 2,3
- use a different way to fix io hung in patch 4
Changes in v3:
- fix a check in patch 1
- fix link err in patch 2 on 32-bit platform
- handle overflow in patch 2
Changes in v2:
- use a new solution suggested by Ming
- change the title of patch 1
- add patch 2
Patch 1 fix that blk-throttle can't work if multiple bios are throttle,
Patch 2 fix overflow while calculating wait time
Patch 3,4 fix io hung due to configuration updates.
Patch 5-8 are cleanup patches, there are no functional changes, just
some places that I think can be optimized during code review.
Previous version:
v1: https://lore.kernel.org/all/[email protected]/
v2: https://lore.kernel.org/all/[email protected]/
v3: https://lore.kernel.org/all/[email protected]/
v4: https://lore.kernel.org/all/[email protected]/
v5: https://lore.kernel.org/all/[email protected]/
Yu Kuai (8):
blk-throttle: fix that io throttle can only work for single bio
blk-throttle: prevent overflow while calculating wait time
blk-throttle: factor out code to calculate ios/bytes_allowed
blk-throttle: fix io hung due to config updates
blk-throttle: use 'READ/WRITE' instead of '0/1'
blk-throttle: calling throtl_dequeue/enqueue_tg in pairs
blk-throttle: cleanup tg_update_disptime()
blk-throttle: clean up flag 'THROTL_TG_PENDING'
block/blk-throttle.c | 168 +++++++++++++++++++++++++++++--------------
block/blk-throttle.h | 16 +++--
2 files changed, 128 insertions(+), 56 deletions(-)
--
2.31.1
All related operations are inside 'queue_lock', there is no need to use
the flag, we only need to make sure throtl_enqueue_tg() is called when
the first bio is throttled, and throtl_dequeue_tg() is called when the
last throttled bio is dispatched.
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 22 ++++++++--------------
block/blk-throttle.h | 7 +++----
2 files changed, 11 insertions(+), 18 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 473f0b651ef0..29e9f7f6573c 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -561,23 +561,16 @@ static void tg_service_queue_add(struct throtl_grp *tg)
static void throtl_enqueue_tg(struct throtl_grp *tg)
{
- if (!(tg->flags & THROTL_TG_PENDING)) {
- tg_service_queue_add(tg);
- tg->flags |= THROTL_TG_PENDING;
- tg->service_queue.parent_sq->nr_pending++;
- }
+ tg_service_queue_add(tg);
+ tg->service_queue.parent_sq->nr_pending++;
}
static void throtl_dequeue_tg(struct throtl_grp *tg)
{
- if (tg->flags & THROTL_TG_PENDING) {
- struct throtl_service_queue *parent_sq =
- tg->service_queue.parent_sq;
+ struct throtl_service_queue *parent_sq = tg->service_queue.parent_sq;
- throtl_rb_erase(&tg->rb_node, parent_sq);
- --parent_sq->nr_pending;
- tg->flags &= ~THROTL_TG_PENDING;
- }
+ throtl_rb_erase(&tg->rb_node, parent_sq);
+ --parent_sq->nr_pending;
}
/* Call with queue lock held */
@@ -1021,8 +1014,9 @@ static void throtl_add_bio_tg(struct bio *bio, struct throtl_qnode *qn,
throtl_qnode_add_bio(bio, qn, &sq->queued[rw]);
+ if (!sq->nr_queued[READ] && !sq->nr_queued[WRITE])
+ throtl_enqueue_tg(tg);
sq->nr_queued[rw]++;
- throtl_enqueue_tg(tg);
}
static void tg_update_disptime(struct throtl_grp *tg)
@@ -1377,7 +1371,7 @@ static void tg_conf_updated(struct throtl_grp *tg, bool global)
throtl_start_new_slice(tg, READ, false);
throtl_start_new_slice(tg, WRITE, false);
- if (tg->flags & THROTL_TG_PENDING) {
+ if (sq->nr_queued[READ] || sq->nr_queued[WRITE]) {
tg_update_disptime(tg);
throtl_schedule_next_dispatch(sq->parent_sq, true);
}
diff --git a/block/blk-throttle.h b/block/blk-throttle.h
index 371d624af845..fba48afbcff3 100644
--- a/block/blk-throttle.h
+++ b/block/blk-throttle.h
@@ -53,10 +53,9 @@ struct throtl_service_queue {
};
enum tg_state_flags {
- THROTL_TG_PENDING = 1 << 0, /* on parent's pending tree */
- THROTL_TG_WAS_EMPTY = 1 << 1, /* bio_lists[] became non-empty */
- THROTL_TG_HAS_IOPS_LIMIT = 1 << 2, /* tg has iops limit */
- THROTL_TG_CANCELING = 1 << 3, /* starts to cancel bio */
+ THROTL_TG_WAS_EMPTY = 1 << 0, /* bio_lists[] became non-empty */
+ THROTL_TG_HAS_IOPS_LIMIT = 1 << 1, /* tg has iops limit */
+ THROTL_TG_CANCELING = 1 << 2, /* starts to cancel bio */
};
enum {
--
2.31.1
Make the code easier to read, like everywhere else.
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 7b09b48577ba..e690dc1c1cde 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -329,8 +329,8 @@ static struct bio *throtl_pop_queued(struct list_head *queued,
/* init a service_queue, assumes the caller zeroed it */
static void throtl_service_queue_init(struct throtl_service_queue *sq)
{
- INIT_LIST_HEAD(&sq->queued[0]);
- INIT_LIST_HEAD(&sq->queued[1]);
+ INIT_LIST_HEAD(&sq->queued[READ]);
+ INIT_LIST_HEAD(&sq->queued[WRITE]);
sq->pending_tree = RB_ROOT_CACHED;
timer_setup(&sq->pending_timer, throtl_pending_timer_fn, 0);
}
@@ -1156,7 +1156,7 @@ static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
nr_disp += throtl_dispatch_tg(tg);
sq = &tg->service_queue;
- if (sq->nr_queued[0] || sq->nr_queued[1])
+ if (sq->nr_queued[READ] || sq->nr_queued[WRITE])
tg_update_disptime(tg);
if (nr_disp >= THROTL_QUANTUM)
--
2.31.1
th_update_disptime() only need to adjust postion for 'tg' in
'parent_sq', there is no need to call throtl_enqueue/dequeue_tg().
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index ab30efedff4e..473f0b651ef0 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -520,7 +520,6 @@ static void throtl_rb_erase(struct rb_node *n,
{
rb_erase_cached(n, &parent_sq->pending_tree);
RB_CLEAR_NODE(n);
- --parent_sq->nr_pending;
}
static void update_min_dispatch_time(struct throtl_service_queue *parent_sq)
@@ -572,7 +571,11 @@ static void throtl_enqueue_tg(struct throtl_grp *tg)
static void throtl_dequeue_tg(struct throtl_grp *tg)
{
if (tg->flags & THROTL_TG_PENDING) {
- throtl_rb_erase(&tg->rb_node, tg->service_queue.parent_sq);
+ struct throtl_service_queue *parent_sq =
+ tg->service_queue.parent_sq;
+
+ throtl_rb_erase(&tg->rb_node, parent_sq);
+ --parent_sq->nr_pending;
tg->flags &= ~THROTL_TG_PENDING;
}
}
@@ -1040,9 +1043,9 @@ static void tg_update_disptime(struct throtl_grp *tg)
disptime = jiffies + min_wait;
/* Update dispatch time */
- throtl_dequeue_tg(tg);
+ throtl_rb_erase(&tg->rb_node, tg->service_queue.parent_sq);
tg->disptime = disptime;
- throtl_enqueue_tg(tg);
+ tg_service_queue_add(tg);
/* see throtl_add_bio_tg() */
tg->flags &= ~THROTL_TG_WAS_EMPTY;
--
2.31.1
In tg_with_in_bps_limit(), 'bps_limit * jiffy_elapsed_rnd' might
overflow. FIx the problem by calling mul_u64_u64_div_u64() instead.
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 8 +++-----
1 file changed, 3 insertions(+), 5 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 5c1d1c4d8188..a89c62bef2fb 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -806,7 +806,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
u64 bps_limit, unsigned long *wait)
{
bool rw = bio_data_dir(bio);
- u64 bytes_allowed, extra_bytes, tmp;
+ u64 bytes_allowed, extra_bytes;
unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
unsigned int bio_size = throtl_bio_data_size(bio);
@@ -824,10 +824,8 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
jiffy_elapsed_rnd = tg->td->throtl_slice;
jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, tg->td->throtl_slice);
-
- tmp = bps_limit * jiffy_elapsed_rnd;
- do_div(tmp, HZ);
- bytes_allowed = tmp;
+ bytes_allowed = mul_u64_u64_div_u64(bps_limit, (u64)jiffy_elapsed_rnd,
+ (u64)HZ);
if (tg->bytes_disp[rw] + bio_size <= bytes_allowed) {
if (wait)
--
2.31.1
If new configuration is submitted while a bio is throttled, then new
waiting time is recalculated regardless that the bio might aready wait
for some time:
tg_conf_updated
throtl_start_new_slice
tg_update_disptime
throtl_schedule_next_dispatch
Then io hung can be triggered by always submmiting new configuration
before the throttled bio is dispatched.
Fix the problem by respecting the time that throttled bio aready waited.
In order to do that, add new fields to record how many bytes/io already
waited, and use it to calculate wait time for throttled bio under new
configuration.
Some simple test:
1)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 2048" > blkio.throttle.write_bps_device
{
sleep 2
echo "8:0 1024" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
2)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 1024" > blkio.throttle.write_bps_device
{
sleep 4
echo "8:0 2048" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
test results: io finish time
before this patch with this patch
1) 10s 6s
2) 8s 6s
Signed-off-by: Yu Kuai <[email protected]>
Reviewed-by: Michal Koutný <[email protected]>
---
block/blk-throttle.c | 58 +++++++++++++++++++++++++++++++++++++++-----
block/blk-throttle.h | 9 +++++++
2 files changed, 61 insertions(+), 6 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 8612a071305e..7b09b48577ba 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -639,6 +639,8 @@ static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
{
tg->bytes_disp[rw] = 0;
tg->io_disp[rw] = 0;
+ tg->bytes_skipped[rw] = 0;
+ tg->io_skipped[rw] = 0;
/*
* Previous slice has expired. We must have trimmed it after last
@@ -656,12 +658,17 @@ static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
tg->slice_end[rw], jiffies);
}
-static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw)
+static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw,
+ bool clear_skipped)
{
tg->bytes_disp[rw] = 0;
tg->io_disp[rw] = 0;
tg->slice_start[rw] = jiffies;
tg->slice_end[rw] = jiffies + tg->td->throtl_slice;
+ if (clear_skipped) {
+ tg->bytes_skipped[rw] = 0;
+ tg->io_skipped[rw] = 0;
+ }
throtl_log(&tg->service_queue,
"[%c] new slice start=%lu end=%lu jiffies=%lu",
@@ -783,6 +790,41 @@ static u64 calculate_bytes_allowed(u64 bps_limit, unsigned long jiffy_elapsed)
return mul_u64_u64_div_u64(bps_limit, (u64)jiffy_elapsed, (u64)HZ);
}
+static void __tg_update_skipped(struct throtl_grp *tg, bool rw)
+{
+ unsigned long jiffy_elapsed = jiffies - tg->slice_start[rw];
+ u64 bps_limit = tg_bps_limit(tg, rw);
+ u32 iops_limit = tg_iops_limit(tg, rw);
+
+ /*
+ * Following calculation won't overflow as long as bios that are
+ * dispatched later won't preempt already throttled bios. Even if such
+ * overflow do happen, there should be no problem because we are using
+ * unsigned here, and bytes_skipped/io_skipped will be updated
+ * correctly.
+ */
+ if (bps_limit != U64_MAX)
+ tg->bytes_skipped[rw] +=
+ calculate_bytes_allowed(bps_limit, jiffy_elapsed) -
+ tg->bytes_disp[rw];
+ if (iops_limit != UINT_MAX)
+ tg->io_skipped[rw] +=
+ calculate_io_allowed(iops_limit, jiffy_elapsed) -
+ tg->io_disp[rw];
+}
+
+static void tg_update_skipped(struct throtl_grp *tg)
+{
+ if (tg->service_queue.nr_queued[READ])
+ __tg_update_skipped(tg, READ);
+ if (tg->service_queue.nr_queued[WRITE])
+ __tg_update_skipped(tg, WRITE);
+
+ throtl_log(&tg->service_queue, "%s: %llu %llu %u %u\n", __func__,
+ tg->bytes_skipped[READ], tg->bytes_skipped[WRITE],
+ tg->io_skipped[READ], tg->io_skipped[WRITE]);
+}
+
static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
u32 iops_limit, unsigned long *wait)
{
@@ -800,7 +842,8 @@ static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
/* Round up to the next throttle slice, wait time must be nonzero */
jiffy_elapsed_rnd = roundup(jiffy_elapsed + 1, tg->td->throtl_slice);
- io_allowed = calculate_io_allowed(iops_limit, jiffy_elapsed_rnd);
+ io_allowed = calculate_io_allowed(iops_limit, jiffy_elapsed_rnd) +
+ tg->io_skipped[rw];
if (tg->io_disp[rw] + 1 <= io_allowed) {
if (wait)
*wait = 0;
@@ -837,7 +880,8 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
jiffy_elapsed_rnd = tg->td->throtl_slice;
jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, tg->td->throtl_slice);
- bytes_allowed = calculate_bytes_allowed(bps_limit, jiffy_elapsed_rnd);
+ bytes_allowed = calculate_bytes_allowed(bps_limit, jiffy_elapsed_rnd) +
+ tg->bytes_skipped[rw];
if (tg->bytes_disp[rw] + bio_size <= bytes_allowed) {
if (wait)
*wait = 0;
@@ -898,7 +942,7 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
* slice and it should be extended instead.
*/
if (throtl_slice_used(tg, rw) && !(tg->service_queue.nr_queued[rw]))
- throtl_start_new_slice(tg, rw);
+ throtl_start_new_slice(tg, rw, true);
else {
if (time_before(tg->slice_end[rw],
jiffies + tg->td->throtl_slice))
@@ -1327,8 +1371,8 @@ static void tg_conf_updated(struct throtl_grp *tg, bool global)
* that a group's limit are dropped suddenly and we don't want to
* account recently dispatched IO with new low rate.
*/
- throtl_start_new_slice(tg, READ);
- throtl_start_new_slice(tg, WRITE);
+ throtl_start_new_slice(tg, READ, false);
+ throtl_start_new_slice(tg, WRITE, false);
if (tg->flags & THROTL_TG_PENDING) {
tg_update_disptime(tg);
@@ -1356,6 +1400,7 @@ static ssize_t tg_set_conf(struct kernfs_open_file *of,
v = U64_MAX;
tg = blkg_to_tg(ctx.blkg);
+ tg_update_skipped(tg);
if (is_u64)
*(u64 *)((void *)tg + of_cft(of)->private) = v;
@@ -1542,6 +1587,7 @@ static ssize_t tg_set_limit(struct kernfs_open_file *of,
return ret;
tg = blkg_to_tg(ctx.blkg);
+ tg_update_skipped(tg);
v[0] = tg->bps_conf[READ][index];
v[1] = tg->bps_conf[WRITE][index];
diff --git a/block/blk-throttle.h b/block/blk-throttle.h
index c1b602996127..371d624af845 100644
--- a/block/blk-throttle.h
+++ b/block/blk-throttle.h
@@ -115,6 +115,15 @@ struct throtl_grp {
uint64_t bytes_disp[2];
/* Number of bio's dispatched in current slice */
unsigned int io_disp[2];
+ /*
+ * The following two fields are used to calculate new wait time for
+ * throttled bio when new configuration is submmited.
+ *
+ * Number of bytes will be skipped in current slice
+ */
+ uint64_t bytes_skipped[2];
+ /* Number of bio will be skipped in current slice */
+ unsigned int io_skipped[2];
unsigned long last_low_overflow_time[2];
--
2.31.1
It's a litter weird to call throtl_dequeue_tg() unconditionally in
throtl_select_dispatch(), since it will be called in
tg_update_disptime() again if some bio is still throttled.
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index e690dc1c1cde..ab30efedff4e 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -1151,13 +1151,13 @@ static int throtl_select_dispatch(struct throtl_service_queue *parent_sq)
if (time_before(jiffies, tg->disptime))
break;
- throtl_dequeue_tg(tg);
-
nr_disp += throtl_dispatch_tg(tg);
sq = &tg->service_queue;
if (sq->nr_queued[READ] || sq->nr_queued[WRITE])
tg_update_disptime(tg);
+ else
+ throtl_dequeue_tg(tg);
if (nr_disp >= THROTL_QUANTUM)
break;
--
2.31.1
If new configuration is submitted while a bio is throttled, then new
waiting time is recalculated regardless that the bio might aready wait
for some time:
tg_conf_updated
throtl_start_new_slice
tg_update_disptime
throtl_schedule_next_dispatch
Then io hung can be triggered by always submmiting new configuration
before the throttled bio is dispatched.
Fix the problem by respecting the time that throttled bio aready waited.
In order to do that, add new fields to record how many bytes/io already
waited, and use it to calculate wait time for throttled bio under new
configuration.
Some simple test:
1)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 2048" > blkio.throttle.write_bps_device
{
sleep 2
echo "8:0 1024" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
2)
cd /sys/fs/cgroup/blkio/
echo $$ > cgroup.procs
echo "8:0 1024" > blkio.throttle.write_bps_device
{
sleep 4
echo "8:0 2048" > blkio.throttle.write_bps_device
} &
dd if=/dev/zero of=/dev/sda bs=8k count=1 oflag=direct
test results: io finish time
before this patch with this patch
1) 10s 6s
2) 8s 6s
Signed-off-by: Yu Kuai <[email protected]>
---
block/blk-throttle.c | 51 ++++++++++++++++++++++++++++++++++++++------
block/blk-throttle.h | 9 ++++++++
2 files changed, 54 insertions(+), 6 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 8612a071305e..8178ef278e7a 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -639,6 +639,8 @@ static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
{
tg->bytes_disp[rw] = 0;
tg->io_disp[rw] = 0;
+ tg->bytes_skipped[rw] = 0;
+ tg->io_skipped[rw] = 0;
/*
* Previous slice has expired. We must have trimmed it after last
@@ -656,12 +658,17 @@ static inline void throtl_start_new_slice_with_credit(struct throtl_grp *tg,
tg->slice_end[rw], jiffies);
}
-static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw)
+static inline void throtl_start_new_slice(struct throtl_grp *tg, bool rw,
+ bool clear_skipped)
{
tg->bytes_disp[rw] = 0;
tg->io_disp[rw] = 0;
tg->slice_start[rw] = jiffies;
tg->slice_end[rw] = jiffies + tg->td->throtl_slice;
+ if (clear_skipped) {
+ tg->bytes_skipped[rw] = 0;
+ tg->io_skipped[rw] = 0;
+ }
throtl_log(&tg->service_queue,
"[%c] new slice start=%lu end=%lu jiffies=%lu",
@@ -783,6 +790,34 @@ static u64 calculate_bytes_allowed(u64 bps_limit, unsigned long jiffy_elapsed)
return mul_u64_u64_div_u64(bps_limit, (u64)jiffy_elapsed, (u64)HZ);
}
+static void __tg_update_skipped(struct throtl_grp *tg, bool rw)
+{
+ unsigned long jiffy_elapsed = jiffies - tg->slice_start[rw];
+ u64 bps_limit = tg_bps_limit(tg, rw);
+ u32 iops_limit = tg_iops_limit(tg, rw);
+
+ if (bps_limit != U64_MAX)
+ tg->bytes_skipped[rw] +=
+ calculate_bytes_allowed(bps_limit, jiffy_elapsed) -
+ tg->bytes_disp[rw];
+ if (iops_limit != UINT_MAX)
+ tg->io_skipped[rw] +=
+ calculate_io_allowed(iops_limit, jiffy_elapsed) -
+ tg->io_disp[rw];
+}
+
+static void tg_update_skipped(struct throtl_grp *tg)
+{
+ if (tg->service_queue.nr_queued[READ])
+ __tg_update_skipped(tg, READ);
+ if (tg->service_queue.nr_queued[WRITE])
+ __tg_update_skipped(tg, WRITE);
+
+ throtl_log(&tg->service_queue, "%s: %llu %llu %u %u\n", __func__,
+ tg->bytes_skipped[READ], tg->bytes_skipped[WRITE],
+ tg->io_skipped[READ], tg->io_skipped[WRITE]);
+}
+
static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
u32 iops_limit, unsigned long *wait)
{
@@ -800,7 +835,8 @@ static bool tg_with_in_iops_limit(struct throtl_grp *tg, struct bio *bio,
/* Round up to the next throttle slice, wait time must be nonzero */
jiffy_elapsed_rnd = roundup(jiffy_elapsed + 1, tg->td->throtl_slice);
- io_allowed = calculate_io_allowed(iops_limit, jiffy_elapsed_rnd);
+ io_allowed = calculate_io_allowed(iops_limit, jiffy_elapsed_rnd) +
+ tg->io_skipped[rw];
if (tg->io_disp[rw] + 1 <= io_allowed) {
if (wait)
*wait = 0;
@@ -837,7 +873,8 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
jiffy_elapsed_rnd = tg->td->throtl_slice;
jiffy_elapsed_rnd = roundup(jiffy_elapsed_rnd, tg->td->throtl_slice);
- bytes_allowed = calculate_bytes_allowed(bps_limit, jiffy_elapsed_rnd);
+ bytes_allowed = calculate_bytes_allowed(bps_limit, jiffy_elapsed_rnd) +
+ tg->bytes_skipped[rw];
if (tg->bytes_disp[rw] + bio_size <= bytes_allowed) {
if (wait)
*wait = 0;
@@ -898,7 +935,7 @@ static bool tg_may_dispatch(struct throtl_grp *tg, struct bio *bio,
* slice and it should be extended instead.
*/
if (throtl_slice_used(tg, rw) && !(tg->service_queue.nr_queued[rw]))
- throtl_start_new_slice(tg, rw);
+ throtl_start_new_slice(tg, rw, true);
else {
if (time_before(tg->slice_end[rw],
jiffies + tg->td->throtl_slice))
@@ -1327,8 +1364,8 @@ static void tg_conf_updated(struct throtl_grp *tg, bool global)
* that a group's limit are dropped suddenly and we don't want to
* account recently dispatched IO with new low rate.
*/
- throtl_start_new_slice(tg, READ);
- throtl_start_new_slice(tg, WRITE);
+ throtl_start_new_slice(tg, READ, false);
+ throtl_start_new_slice(tg, WRITE, false);
if (tg->flags & THROTL_TG_PENDING) {
tg_update_disptime(tg);
@@ -1356,6 +1393,7 @@ static ssize_t tg_set_conf(struct kernfs_open_file *of,
v = U64_MAX;
tg = blkg_to_tg(ctx.blkg);
+ tg_update_skipped(tg);
if (is_u64)
*(u64 *)((void *)tg + of_cft(of)->private) = v;
@@ -1542,6 +1580,7 @@ static ssize_t tg_set_limit(struct kernfs_open_file *of,
return ret;
tg = blkg_to_tg(ctx.blkg);
+ tg_update_skipped(tg);
v[0] = tg->bps_conf[READ][index];
v[1] = tg->bps_conf[WRITE][index];
diff --git a/block/blk-throttle.h b/block/blk-throttle.h
index c1b602996127..371d624af845 100644
--- a/block/blk-throttle.h
+++ b/block/blk-throttle.h
@@ -115,6 +115,15 @@ struct throtl_grp {
uint64_t bytes_disp[2];
/* Number of bio's dispatched in current slice */
unsigned int io_disp[2];
+ /*
+ * The following two fields are used to calculate new wait time for
+ * throttled bio when new configuration is submmited.
+ *
+ * Number of bytes will be skipped in current slice
+ */
+ uint64_t bytes_skipped[2];
+ /* Number of bio will be skipped in current slice */
+ unsigned int io_skipped[2];
unsigned long last_low_overflow_time[2];
--
2.31.1
commit 9f5ede3c01f9 ("block: throttle split bio in case of iops limit")
introduce a new problem, for example:
Test scripts:
cd /sys/fs/cgroup/blkio/
echo "8:0 1024" > blkio.throttle.write_bps_device
echo $$ > cgroup.procs
dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
Test result:
10240 bytes (10 kB, 10 KiB) copied, 10.0134 s, 1.0 kB/s
10240 bytes (10 kB, 10 KiB) copied, 10.0135 s, 1.0 kB/s
The problem is that the second bio is finished after 10s instead of 20s.
This is because if some bios are already queued, current bio is queued
directly and the flag 'BIO_THROTTLED' is set. And later, when former
bios are dispatched, this bio will be dispatched without waiting at all,
this is due to tg_with_in_bps_limit() return 0 for this bio.
In order to fix the problem, don't skip flaged bio in
tg_with_in_bps_limit(), and for the problem that split bio can be
double accounted, compensate the over-accounting in __blk_throtl_bio().
Fixes: 9f5ede3c01f9 ("block: throttle split bio in case of iops limit")
Signed-off-by: Yu Kuai <[email protected]>
Reviewed-by: Ming Lei <[email protected]>
---
block/blk-throttle.c | 24 ++++++++++++++++++------
1 file changed, 18 insertions(+), 6 deletions(-)
diff --git a/block/blk-throttle.c b/block/blk-throttle.c
index 139b2d7a99e2..5c1d1c4d8188 100644
--- a/block/blk-throttle.c
+++ b/block/blk-throttle.c
@@ -811,7 +811,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
unsigned int bio_size = throtl_bio_data_size(bio);
/* no need to throttle if this bio's bytes have been accounted */
- if (bps_limit == U64_MAX || bio_flagged(bio, BIO_THROTTLED)) {
+ if (bps_limit == U64_MAX) {
if (wait)
*wait = 0;
return true;
@@ -921,11 +921,8 @@ static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
unsigned int bio_size = throtl_bio_data_size(bio);
/* Charge the bio to the group */
- if (!bio_flagged(bio, BIO_THROTTLED)) {
- tg->bytes_disp[rw] += bio_size;
- tg->last_bytes_disp[rw] += bio_size;
- }
-
+ tg->bytes_disp[rw] += bio_size;
+ tg->last_bytes_disp[rw] += bio_size;
tg->io_disp[rw]++;
tg->last_io_disp[rw]++;
@@ -2121,6 +2118,21 @@ bool __blk_throtl_bio(struct bio *bio)
tg->last_low_overflow_time[rw] = jiffies;
throtl_downgrade_check(tg);
throtl_upgrade_check(tg);
+
+ /*
+ * re-entered bio has accounted bytes already, so try to
+ * compensate previous over-accounting. However, if new
+ * slice is started, just forget it.
+ */
+ if (bio_flagged(bio, BIO_THROTTLED)) {
+ unsigned int bio_size = throtl_bio_data_size(bio);
+
+ if (tg->bytes_disp[rw] >= bio_size)
+ tg->bytes_disp[rw] -= bio_size;
+ if (tg->last_bytes_disp[rw] >= bio_size)
+ tg->last_bytes_disp[rw] -= bio_size;
+ }
+
/* throtl is FIFO - if bios are already queued, should queue */
if (sq->nr_queued[rw])
break;
--
2.31.1