mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-26 10:35:04 +07:00
76758390f8
In a SSD, write amplification, WA, is defined as the average
number of page writes per user page write. Write amplification
negatively affects write performance and decreases the lifetime
of the disk, so it's a useful metric to add to sysfs.
In plkb's case, the number of writes per user sector is the sum of:
(1) number of user writes
(2) number of sectors written by the garbage collector
(3) number of sectors padded (i.e. due to syncs)
This patch adds persistent counters for 1-3 and two sysfs attributes
to export these along with WA calculated with five decimals:
write_amp_mileage: the accumulated write amplification stats
for the lifetime of the pblk instance
write_amp_trip: resetable stats to facilitate delta measurements,
values reset at creation and if 0 is written
to the attribute.
64-bit counters are used as a 32 bit counter would wrap around
already after about 17 TB worth of user data. It will take a
long long time before the 64 bit sector counters wrap around.
The counters are stored after the bad block bitmap in the first
emeta sector of each written line. There is plenty of space in the
first emeta sector, so we don't need to bump the major version of
the line data format.
Signed-off-by: Hans Holmberg <hans.holmberg@cnexlabs.com>
Signed-off-by: Javier González <javier@cnexlabs.com>
Signed-off-by: Matias Bjørling <mb@lightnvm.io>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
132 lines
3.4 KiB
C
132 lines
3.4 KiB
C
/*
|
|
* Copyright (C) 2016 CNEX Labs
|
|
* Initial release: Javier Gonzalez <javier@cnexlabs.com>
|
|
* Matias Bjorling <matias@cnexlabs.com>
|
|
*
|
|
* This program is free software; you can redistribute it and/or
|
|
* modify it under the terms of the GNU General Public License version
|
|
* 2 as published by the Free Software Foundation.
|
|
*
|
|
* This program is distributed in the hope that it will be useful, but
|
|
* WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
* General Public License for more details.
|
|
*
|
|
* pblk-cache.c - pblk's write cache
|
|
*/
|
|
|
|
#include "pblk.h"
|
|
|
|
int pblk_write_to_cache(struct pblk *pblk, struct bio *bio, unsigned long flags)
|
|
{
|
|
struct request_queue *q = pblk->dev->q;
|
|
struct pblk_w_ctx w_ctx;
|
|
sector_t lba = pblk_get_lba(bio);
|
|
unsigned long start_time = jiffies;
|
|
unsigned int bpos, pos;
|
|
int nr_entries = pblk_get_secs(bio);
|
|
int i, ret;
|
|
|
|
generic_start_io_acct(q, WRITE, bio_sectors(bio), &pblk->disk->part0);
|
|
|
|
/* Update the write buffer head (mem) with the entries that we can
|
|
* write. The write in itself cannot fail, so there is no need to
|
|
* rollback from here on.
|
|
*/
|
|
retry:
|
|
ret = pblk_rb_may_write_user(&pblk->rwb, bio, nr_entries, &bpos);
|
|
switch (ret) {
|
|
case NVM_IO_REQUEUE:
|
|
io_schedule();
|
|
goto retry;
|
|
case NVM_IO_ERR:
|
|
pblk_pipeline_stop(pblk);
|
|
goto out;
|
|
}
|
|
|
|
if (unlikely(!bio_has_data(bio)))
|
|
goto out;
|
|
|
|
pblk_ppa_set_empty(&w_ctx.ppa);
|
|
w_ctx.flags = flags;
|
|
if (bio->bi_opf & REQ_PREFLUSH)
|
|
w_ctx.flags |= PBLK_FLUSH_ENTRY;
|
|
|
|
for (i = 0; i < nr_entries; i++) {
|
|
void *data = bio_data(bio);
|
|
|
|
w_ctx.lba = lba + i;
|
|
|
|
pos = pblk_rb_wrap_pos(&pblk->rwb, bpos + i);
|
|
pblk_rb_write_entry_user(&pblk->rwb, data, w_ctx, pos);
|
|
|
|
bio_advance(bio, PBLK_EXPOSED_PAGE_SIZE);
|
|
}
|
|
|
|
atomic64_add(nr_entries, &pblk->user_wa);
|
|
|
|
#ifdef CONFIG_NVM_DEBUG
|
|
atomic_long_add(nr_entries, &pblk->inflight_writes);
|
|
atomic_long_add(nr_entries, &pblk->req_writes);
|
|
#endif
|
|
|
|
pblk_rl_inserted(&pblk->rl, nr_entries);
|
|
|
|
out:
|
|
generic_end_io_acct(q, WRITE, &pblk->disk->part0, start_time);
|
|
pblk_write_should_kick(pblk);
|
|
return ret;
|
|
}
|
|
|
|
/*
|
|
* On GC the incoming lbas are not necessarily sequential. Also, some of the
|
|
* lbas might not be valid entries, which are marked as empty by the GC thread
|
|
*/
|
|
int pblk_write_gc_to_cache(struct pblk *pblk, struct pblk_gc_rq *gc_rq)
|
|
{
|
|
struct pblk_w_ctx w_ctx;
|
|
unsigned int bpos, pos;
|
|
void *data = gc_rq->data;
|
|
int i, valid_entries;
|
|
|
|
/* Update the write buffer head (mem) with the entries that we can
|
|
* write. The write in itself cannot fail, so there is no need to
|
|
* rollback from here on.
|
|
*/
|
|
retry:
|
|
if (!pblk_rb_may_write_gc(&pblk->rwb, gc_rq->secs_to_gc, &bpos)) {
|
|
io_schedule();
|
|
goto retry;
|
|
}
|
|
|
|
w_ctx.flags = PBLK_IOTYPE_GC;
|
|
pblk_ppa_set_empty(&w_ctx.ppa);
|
|
|
|
for (i = 0, valid_entries = 0; i < gc_rq->nr_secs; i++) {
|
|
if (gc_rq->lba_list[i] == ADDR_EMPTY)
|
|
continue;
|
|
|
|
w_ctx.lba = gc_rq->lba_list[i];
|
|
|
|
pos = pblk_rb_wrap_pos(&pblk->rwb, bpos + valid_entries);
|
|
pblk_rb_write_entry_gc(&pblk->rwb, data, w_ctx, gc_rq->line,
|
|
gc_rq->paddr_list[i], pos);
|
|
|
|
data += PBLK_EXPOSED_PAGE_SIZE;
|
|
valid_entries++;
|
|
}
|
|
|
|
WARN_ONCE(gc_rq->secs_to_gc != valid_entries,
|
|
"pblk: inconsistent GC write\n");
|
|
|
|
atomic64_add(valid_entries, &pblk->gc_wa);
|
|
|
|
#ifdef CONFIG_NVM_DEBUG
|
|
atomic_long_add(valid_entries, &pblk->inflight_writes);
|
|
atomic_long_add(valid_entries, &pblk->recov_gc_writes);
|
|
#endif
|
|
|
|
pblk_write_should_kick(pblk);
|
|
return NVM_IO_OK;
|
|
}
|