[Devel] [PATCH RHEL9 COMMIT] dm-ploop: fast path execution for reads
Konstantin Khorenko
khorenko at virtuozzo.com
Mon Jan 27 16:12:36 MSK 2025
The commit is pushed to "branch-rh9-5.14.0-427.44.1.vz9.80.x-ovz" and will appear at git at bitbucket.org:openvz/vzkernel.git
after rh9-5.14.0-427.44.1.vz9.80.6
------>
commit 46382ef09dc5d44f3c440f05fa826d25a1da5545
Author: Alexander Atanasov <alexander.atanasov at virtuozzo.com>
Date: Fri Jan 24 17:35:46 2025 +0200
dm-ploop: fast path execution for reads
After a pio is split and prepared try to execute if immediately
without going to the worker thread.
https://virtuozzo.atlassian.net/browse/VSTOR-91820
Signed-off-by: Alexander Atanasov <alexander.atanasov at virtuozzo.com>
======
Patchset description:
ploop: optimistations and scalling
Ploop processes requsts in a different threads in parallel
where possible which results in significant improvement in
performance and makes further optimistations possible.
Known bugs:
- delayed metadata writeback is not working and is missing error handling
- patch to disable it until fixed
- fast path is not working - causes rcu lockups - patch to disable it
Further improvements:
- optimize md pages lookups
Alexander Atanasov (50):
dm-ploop: md_pages map all pages at creation time
dm-ploop: Use READ_ONCE/WRITE_ONCE to access md page data
dm-ploop: fsync after all pios are sent
dm-ploop: move md status to use proper bitops
dm-ploop: convert wait_list and wb_batch_llist to use lockless lists
dm-ploop: convert enospc handling to use lockless lists
dm-ploop: convert suspended_pios list to use lockless list
dm-ploop: convert the rest of the lists to use llist variant
dm-ploop: combine processing of pios thru prepare list and remove
fsync worker
dm-ploop: move from wq to kthread
dm-ploop: move preparations of pios into the caller from worker
dm-ploop: fast path execution for reads
dm-ploop: do not use a wrapper for set_bit to make a page writeback
dm-ploop: BAT use only one list for writeback
dm-ploop: make md writeback timeout to be per page
dm-ploop: add interface to disable bat writeback delay
dm-ploop: convert wb_batch_list to lockless variant
dm-ploop: convert high_prio to status
dm-ploop: split cow processing into two functions
dm-ploop: convert md page rw lock to spin lock
dm-ploop: convert bat_rwlock to bat_lock spinlock
dm-ploop: prepare bat updates under bat_lock
dm-ploop: make ploop_bat_write_complete ready for parallel pio
completion
dm-ploop: make ploop_submit_metadata_writeback return number of
requests sent
dm-ploop: introduce pio runner threads
dm-ploop: add pio list ids to be used when passing pios to runners
dm-ploop: process pios via runners
dm-ploop: disable metadata writeback delay
dm-ploop: disable fast path
dm-ploop: use lockless lists for chained cow updates list
dm-ploop: use lockless lists for data ready pios
dm-ploop: give runner threads better name
dm-ploop: resize operation - add holes bitmap locking
dm-ploop: remove unnecessary operations
dm-ploop: use filp per thread
dm-ploop: catch if we try to advance pio past bio end
dm-ploop: support REQ_FUA for data pios
dm-ploop: proplerly access nr_bat_entries
dm-ploop: fix locking and improve error handling when submitting pios
dm-ploop: fix how ENOTBLK is handled
dm-ploop: sync when suspended or stopping
dm-ploop: rework bat completion logic
dm-ploop: rework logic in pio processing
dm-ploop: end fsync pios in parallel
dm-ploop: make filespace preallocations async
dm-ploop: resubmit enospc pios from dispatcher thread
dm-ploop: dm-ploop: simplify discard completion
dm-ploop: use GFP_ATOMIC instead of GFP_NOIO
dm-ploop: fix locks used in mixed context
dm-ploop: fix how current flags are managed inside threads
Andrey Zhadchenko (13):
dm-ploop: do not flush after metadata writes
dm-ploop: set IOCB_DSYNC on all FUA requests
dm-ploop: remove extra ploop_cluster_is_in_top_delta()
dm-ploop: introduce per-md page locking
dm-ploop: reduce BAT accesses on discard completion
dm-ploop: simplify llseek
dm-ploop: speed up ploop_prepare_bat_update()
dm-ploop: make new allocations immediately visible in BAT
dm-ploop: drop ploop_cluster_is_in_top_delta()
dm-ploop: do not wait for BAT update for non-FUA requests
dm-ploop: add delay for metadata writeback
dm-ploop: submit all postponed metadata on REQ_OP_FLUSH
dm-ploop: handle REQ_PREFLUSH
Feature: dm-ploop: ploop target driver
---
drivers/md/dm-ploop-map.c | 96 ++++++++++++++++++++++++++++++++++-------------
1 file changed, 70 insertions(+), 26 deletions(-)
diff --git a/drivers/md/dm-ploop-map.c b/drivers/md/dm-ploop-map.c
index 897b4cb04746..8e1d4bf6438a 100644
--- a/drivers/md/dm-ploop-map.c
+++ b/drivers/md/dm-ploop-map.c
@@ -604,7 +604,7 @@ static int ploop_zero_range(struct file *file, loff_t pos, loff_t len)
pos, len);
}
-static void ploop_handle_discard_pio(struct ploop *ploop, struct pio *pio,
+static int ploop_handle_discard_pio(struct ploop *ploop, struct pio *pio,
u32 clu, u32 dst_clu)
{
struct pio *inflight_h;
@@ -621,12 +621,12 @@ static void ploop_handle_discard_pio(struct ploop *ploop, struct pio *pio,
* this implementing REQ_OP_WRITE_ZEROES etc.
*/
ploop_pio_endio(pio);
- return;
+ return 1;
}
if (!ploop_cluster_is_in_top_delta(ploop, clu)) {
ploop_pio_endio(pio);
- return;
+ return 1;
}
/* We can't end with EOPNOTSUPP, since blk-mq prints error */
@@ -643,7 +643,7 @@ static void ploop_handle_discard_pio(struct ploop *ploop, struct pio *pio,
/* @pio will be requeued on inflight_h's pio end */
pr_err_once(PL_FMT("delayed discard: device is used as raw?"),
ploop_device_name(ploop));
- return;
+ return 0;
}
ploop_add_cluster_lk(ploop, pio, clu);
@@ -658,10 +658,11 @@ static void ploop_handle_discard_pio(struct ploop *ploop, struct pio *pio,
if (ret)
pio->bi_status = errno_to_blk_status(ret);
ploop_pio_endio(pio);
- return;
+ return 1;
}
ploop_queue_discard_index_wb(ploop, pio);
+ return 0;
}
static void ploop_discard_index_pio_end(struct ploop *ploop, struct pio *pio)
@@ -1257,17 +1258,18 @@ void ploop_map_and_submit_rw(struct ploop *ploop, u32 dst_clu,
ploop_submit_rw_mapped(ploop, pio);
}
-static void ploop_initiate_delta_read(struct ploop *ploop, unsigned int level,
+static int ploop_initiate_delta_read(struct ploop *ploop, unsigned int level,
u32 dst_clu, struct pio *pio)
{
if (dst_clu == BAT_ENTRY_NONE) {
/* No one delta contains dst_clu. */
ploop_zero_fill_pio(pio);
ploop_pio_endio(pio);
- return;
+ return 1;
}
ploop_map_and_submit_rw(ploop, dst_clu, pio, level);
+ return 0;
}
static void ploop_cow_endio(struct pio *aux_pio, void *data,
@@ -1328,14 +1330,15 @@ static int ploop_submit_cluster_cow(struct ploop *ploop, unsigned int level,
return -ENOMEM;
}
-static void ploop_initiate_cluster_cow(struct ploop *ploop, unsigned int level,
+static int ploop_initiate_cluster_cow(struct ploop *ploop, unsigned int level,
u32 clu, u32 dst_clu, struct pio *pio)
{
if (!ploop_submit_cluster_cow(ploop, level, clu, dst_clu, pio))
- return;
+ return 0;
pio->bi_status = BLK_STS_RESOURCE;
ploop_pio_endio(pio);
+ return 1;
}
ALLOW_ERROR_INJECTION(ploop_submit_cluster_cow, ERRNO);
@@ -1510,23 +1513,26 @@ static int ploop_process_one_deferred_bio(struct ploop *ploop, struct pio *pio)
struct md_page *md;
u32 clu, dst_clu;
u8 level;
- bool ret;
+ int ret;
clu = SEC_TO_CLU(ploop, sector);
+ /* Pio is put into endio_list */
if (ploop_postpone_if_cluster_locked(ploop, pio, clu))
- goto out;
+ goto handled;
dst_clu = ploop_bat_entries(ploop, clu, &level, &md);
if (op_is_discard(pio->bi_op)) {
/* FIXME: check there is no parallel alloc */
- ploop_handle_discard_pio(ploop, pio, clu, dst_clu);
- goto out;
+ ret = ploop_handle_discard_pio(ploop, pio, clu, dst_clu);
+ if (ret)
+ goto handled;
+ goto executed;
}
if (ploop_cluster_is_in_top_delta(ploop, clu)) {
/* Already mapped */
if (ploop_pio_endio_if_merge_fake_pio(pio))
- goto out;
+ goto handled;
goto queue;
} else if (!op_is_write(pio->bi_op)) {
/*
@@ -1534,7 +1540,7 @@ static int ploop_process_one_deferred_bio(struct ploop *ploop, struct pio *pio)
* (Also handles the case dst_clu == BAT_ENTRY_NONE).
*/
ploop_initiate_delta_read(ploop, level, dst_clu, pio);
- goto out;
+ goto executed;
} else if (dst_clu != BAT_ENTRY_NONE) {
/*
* Read secondary delta and write to top delta. May fail.
@@ -1542,23 +1548,27 @@ static int ploop_process_one_deferred_bio(struct ploop *ploop, struct pio *pio)
* a lot of other corner cases, but we don't do that as
* snapshots are used and COW occurs very rare.
*/
- ploop_initiate_cluster_cow(ploop, level, clu, dst_clu, pio);
- goto out;
+ if (!ploop_initiate_cluster_cow(ploop, level, clu, dst_clu, pio))
+ goto executed;
+ goto handled;
}
+ /* pio is completed on success */
if (unlikely(ploop_pio_endio_if_all_zeros(pio)))
- goto out;
+ goto handled;
/* Cluster exists nowhere. Allocate it and setup pio as outrunning */
ret = ploop_locate_new_cluster_and_attach_pio(ploop, md, clu,
&dst_clu, pio);
if (!ret)
- goto out;
-queue:
+ goto executed;
+queue: /* pio needs to go to the queue */
ploop_link_submitting_pio(ploop, pio, clu);
-
ploop_map_and_submit_rw(ploop, dst_clu, pio, ploop_top_level(ploop));
-out:
+ // return 1;
+executed: /* pio op started */
+ return 2;
+handled: /* do not touch pio is completed */
return 0;
}
@@ -1924,7 +1934,7 @@ static void ploop_submit_embedded_pio(struct ploop *ploop, struct pio *pio)
struct request *rq = prq->rq;
LLIST_HEAD(deferred_pios);
struct pio *spio;
- struct llist_node *pos, *t;
+ int ret;
if (blk_rq_bytes(rq)) {
pio->queue_list_id = PLOOP_LIST_PREPARE;
@@ -1939,11 +1949,45 @@ static void ploop_submit_embedded_pio(struct ploop *ploop, struct pio *pio)
}
ploop_inc_nr_inflight(ploop, pio);
+
+ if (pio->queue_list_id == PLOOP_LIST_FLUSH) {
+ /* Let the FLUSH go last from the queue , do not run here to preserve ordering */
+ llist_add((struct llist_node *)(&pio->list), &ploop->pios[PLOOP_LIST_PREPARE]);
+ return;
+ }
+
ploop_prepare_one_embedded_pio(ploop, pio, &deferred_pios);
+ /* check if result is a single pio , so we can try fast path */
+ if (deferred_pios.first && deferred_pios.first->next == NULL) {
+ spio = list_entry((struct list_head *)deferred_pios.first, typeof(*spio), list);
+ INIT_LIST_HEAD(&spio->list);
+ if (req_op(rq) == REQ_OP_READ) {
+ unsigned int old_flags = current->flags;
+
+ current->flags |= PF_IO_THREAD|PF_LOCAL_THROTTLE|PF_MEMALLOC_NOIO;
+ ret = ploop_process_one_deferred_bio(ploop, spio);
+ current->flags = old_flags;
+ if (ret == 1) {
+ /* not queued add back to deferreed*/
+ llist_add((struct llist_node *)(&spio->list),
+ &ploop->pios[PLOOP_LIST_DEFERRED]);
+ } else {
+ return;
+ }
+
+ } else { /* OP which can not go thru fast path */
+ INIT_LIST_HEAD(&spio->list);
+ llist_add((struct llist_node *)(&spio->list),
+ &ploop->pios[PLOOP_LIST_DEFERRED]);
+ }
+ } else {
+ struct llist_node *pos, *t;
- llist_for_each_safe(pos, t, deferred_pios.first) {
- spio = list_entry((struct list_head *)pos, typeof(*pio), list);
- llist_add((struct llist_node *)(&spio->list), &ploop->pios[PLOOP_LIST_DEFERRED]);
+ llist_for_each_safe(pos, t, deferred_pios.first) {
+ spio = list_entry((struct list_head *)pos, typeof(*pio), list);
+ llist_add((struct llist_node *)(&spio->list),
+ &ploop->pios[PLOOP_LIST_DEFERRED]);
+ }
}
ploop_schedule_work(ploop);
More information about the Devel
mailing list