From: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
To: linux-kernel@vger.kernel.org
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
stable@vger.kernel.org, Light Hsieh <Light.Hsieh@mediatek.com>,
Chao Yu <yuchao0@huawei.com>, Jaegeuk Kim <jaegeuk@kernel.org>,
Sasha Levin <sashal@kernel.org>
Subject: [PATCH 5.10 46/63] f2fs: avoid race condition for shrinker count
Date: Mon, 4 Jan 2021 16:57:39 +0100 [thread overview]
Message-ID: <20210104155711.049210727@linuxfoundation.org> (raw)
In-Reply-To: <20210104155708.800470590@linuxfoundation.org>
From: Jaegeuk Kim <jaegeuk@kernel.org>
[ Upstream commit a95ba66ac1457b76fe472c8e092ab1006271f16c ]
Light reported sometimes shinker gets nat_cnt < dirty_nat_cnt resulting in
wrong do_shinker work. Let's avoid to return insane overflowed value by adding
single tracking value.
Reported-by: Light Hsieh <Light.Hsieh@mediatek.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
---
fs/f2fs/checkpoint.c | 2 +-
fs/f2fs/debug.c | 11 ++++++-----
fs/f2fs/f2fs.h | 10 ++++++++--
fs/f2fs/node.c | 29 ++++++++++++++++++-----------
fs/f2fs/node.h | 4 ++--
fs/f2fs/shrinker.c | 4 +---
6 files changed, 36 insertions(+), 24 deletions(-)
diff --git a/fs/f2fs/checkpoint.c b/fs/f2fs/checkpoint.c
index 023462e80e58d..b39bf416d5114 100644
--- a/fs/f2fs/checkpoint.c
+++ b/fs/f2fs/checkpoint.c
@@ -1600,7 +1600,7 @@ int f2fs_write_checkpoint(struct f2fs_sb_info *sbi, struct cp_control *cpc)
goto out;
}
- if (NM_I(sbi)->dirty_nat_cnt == 0 &&
+ if (NM_I(sbi)->nat_cnt[DIRTY_NAT] == 0 &&
SIT_I(sbi)->dirty_sentries == 0 &&
prefree_segments(sbi) == 0) {
f2fs_flush_sit_entries(sbi, cpc);
diff --git a/fs/f2fs/debug.c b/fs/f2fs/debug.c
index a8357fd4f5fab..197c914119da8 100644
--- a/fs/f2fs/debug.c
+++ b/fs/f2fs/debug.c
@@ -145,8 +145,8 @@ static void update_general_status(struct f2fs_sb_info *sbi)
si->node_pages = NODE_MAPPING(sbi)->nrpages;
if (sbi->meta_inode)
si->meta_pages = META_MAPPING(sbi)->nrpages;
- si->nats = NM_I(sbi)->nat_cnt;
- si->dirty_nats = NM_I(sbi)->dirty_nat_cnt;
+ si->nats = NM_I(sbi)->nat_cnt[TOTAL_NAT];
+ si->dirty_nats = NM_I(sbi)->nat_cnt[DIRTY_NAT];
si->sits = MAIN_SEGS(sbi);
si->dirty_sits = SIT_I(sbi)->dirty_sentries;
si->free_nids = NM_I(sbi)->nid_cnt[FREE_NID];
@@ -278,9 +278,10 @@ static void update_mem_info(struct f2fs_sb_info *sbi)
si->cache_mem += (NM_I(sbi)->nid_cnt[FREE_NID] +
NM_I(sbi)->nid_cnt[PREALLOC_NID]) *
sizeof(struct free_nid);
- si->cache_mem += NM_I(sbi)->nat_cnt * sizeof(struct nat_entry);
- si->cache_mem += NM_I(sbi)->dirty_nat_cnt *
- sizeof(struct nat_entry_set);
+ si->cache_mem += NM_I(sbi)->nat_cnt[TOTAL_NAT] *
+ sizeof(struct nat_entry);
+ si->cache_mem += NM_I(sbi)->nat_cnt[DIRTY_NAT] *
+ sizeof(struct nat_entry_set);
si->cache_mem += si->inmem_pages * sizeof(struct inmem_pages);
for (i = 0; i < MAX_INO_ENTRY; i++)
si->cache_mem += sbi->im[i].ino_num * sizeof(struct ino_entry);
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
index 9a321c52facec..e4344d98a780c 100644
--- a/fs/f2fs/f2fs.h
+++ b/fs/f2fs/f2fs.h
@@ -894,6 +894,13 @@ enum nid_state {
MAX_NID_STATE,
};
+enum nat_state {
+ TOTAL_NAT,
+ DIRTY_NAT,
+ RECLAIMABLE_NAT,
+ MAX_NAT_STATE,
+};
+
struct f2fs_nm_info {
block_t nat_blkaddr; /* base disk address of NAT */
nid_t max_nid; /* maximum possible node ids */
@@ -909,8 +916,7 @@ struct f2fs_nm_info {
struct rw_semaphore nat_tree_lock; /* protect nat_tree_lock */
struct list_head nat_entries; /* cached nat entry list (clean) */
spinlock_t nat_list_lock; /* protect clean nat entry list */
- unsigned int nat_cnt; /* the # of cached nat entries */
- unsigned int dirty_nat_cnt; /* total num of nat entries in set */
+ unsigned int nat_cnt[MAX_NAT_STATE]; /* the # of cached nat entries */
unsigned int nat_blocks; /* # of nat blocks */
/* free node ids management */
diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
index 42394de6c7eb1..e65d73293a3f6 100644
--- a/fs/f2fs/node.c
+++ b/fs/f2fs/node.c
@@ -62,8 +62,8 @@ bool f2fs_available_free_memory(struct f2fs_sb_info *sbi, int type)
sizeof(struct free_nid)) >> PAGE_SHIFT;
res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
} else if (type == NAT_ENTRIES) {
- mem_size = (nm_i->nat_cnt * sizeof(struct nat_entry)) >>
- PAGE_SHIFT;
+ mem_size = (nm_i->nat_cnt[TOTAL_NAT] *
+ sizeof(struct nat_entry)) >> PAGE_SHIFT;
res = mem_size < ((avail_ram * nm_i->ram_thresh / 100) >> 2);
if (excess_cached_nats(sbi))
res = false;
@@ -177,7 +177,8 @@ static struct nat_entry *__init_nat_entry(struct f2fs_nm_info *nm_i,
list_add_tail(&ne->list, &nm_i->nat_entries);
spin_unlock(&nm_i->nat_list_lock);
- nm_i->nat_cnt++;
+ nm_i->nat_cnt[TOTAL_NAT]++;
+ nm_i->nat_cnt[RECLAIMABLE_NAT]++;
return ne;
}
@@ -207,7 +208,8 @@ static unsigned int __gang_lookup_nat_cache(struct f2fs_nm_info *nm_i,
static void __del_from_nat_cache(struct f2fs_nm_info *nm_i, struct nat_entry *e)
{
radix_tree_delete(&nm_i->nat_root, nat_get_nid(e));
- nm_i->nat_cnt--;
+ nm_i->nat_cnt[TOTAL_NAT]--;
+ nm_i->nat_cnt[RECLAIMABLE_NAT]--;
__free_nat_entry(e);
}
@@ -253,7 +255,8 @@ static void __set_nat_cache_dirty(struct f2fs_nm_info *nm_i,
if (get_nat_flag(ne, IS_DIRTY))
goto refresh_list;
- nm_i->dirty_nat_cnt++;
+ nm_i->nat_cnt[DIRTY_NAT]++;
+ nm_i->nat_cnt[RECLAIMABLE_NAT]--;
set_nat_flag(ne, IS_DIRTY, true);
refresh_list:
spin_lock(&nm_i->nat_list_lock);
@@ -273,7 +276,8 @@ static void __clear_nat_cache_dirty(struct f2fs_nm_info *nm_i,
set_nat_flag(ne, IS_DIRTY, false);
set->entry_cnt--;
- nm_i->dirty_nat_cnt--;
+ nm_i->nat_cnt[DIRTY_NAT]--;
+ nm_i->nat_cnt[RECLAIMABLE_NAT]++;
}
static unsigned int __gang_lookup_nat_set(struct f2fs_nm_info *nm_i,
@@ -2944,14 +2948,17 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
LIST_HEAD(sets);
int err = 0;
- /* during unmount, let's flush nat_bits before checking dirty_nat_cnt */
+ /*
+ * during unmount, let's flush nat_bits before checking
+ * nat_cnt[DIRTY_NAT].
+ */
if (enabled_nat_bits(sbi, cpc)) {
down_write(&nm_i->nat_tree_lock);
remove_nats_in_journal(sbi);
up_write(&nm_i->nat_tree_lock);
}
- if (!nm_i->dirty_nat_cnt)
+ if (!nm_i->nat_cnt[DIRTY_NAT])
return 0;
down_write(&nm_i->nat_tree_lock);
@@ -2962,7 +2969,8 @@ int f2fs_flush_nat_entries(struct f2fs_sb_info *sbi, struct cp_control *cpc)
* into nat entry set.
*/
if (enabled_nat_bits(sbi, cpc) ||
- !__has_cursum_space(journal, nm_i->dirty_nat_cnt, NAT_JOURNAL))
+ !__has_cursum_space(journal,
+ nm_i->nat_cnt[DIRTY_NAT], NAT_JOURNAL))
remove_nats_in_journal(sbi);
while ((found = __gang_lookup_nat_set(nm_i,
@@ -3086,7 +3094,6 @@ static int init_node_manager(struct f2fs_sb_info *sbi)
F2FS_RESERVED_NODE_NUM;
nm_i->nid_cnt[FREE_NID] = 0;
nm_i->nid_cnt[PREALLOC_NID] = 0;
- nm_i->nat_cnt = 0;
nm_i->ram_thresh = DEF_RAM_THRESHOLD;
nm_i->ra_nid_pages = DEF_RA_NID_PAGES;
nm_i->dirty_nats_ratio = DEF_DIRTY_NAT_RATIO_THRESHOLD;
@@ -3220,7 +3227,7 @@ void f2fs_destroy_node_manager(struct f2fs_sb_info *sbi)
__del_from_nat_cache(nm_i, natvec[idx]);
}
}
- f2fs_bug_on(sbi, nm_i->nat_cnt);
+ f2fs_bug_on(sbi, nm_i->nat_cnt[TOTAL_NAT]);
/* destroy nat set cache */
nid = 0;
diff --git a/fs/f2fs/node.h b/fs/f2fs/node.h
index 69e5859e993cf..f84541b57acbb 100644
--- a/fs/f2fs/node.h
+++ b/fs/f2fs/node.h
@@ -126,13 +126,13 @@ static inline void raw_nat_from_node_info(struct f2fs_nat_entry *raw_ne,
static inline bool excess_dirty_nats(struct f2fs_sb_info *sbi)
{
- return NM_I(sbi)->dirty_nat_cnt >= NM_I(sbi)->max_nid *
+ return NM_I(sbi)->nat_cnt[DIRTY_NAT] >= NM_I(sbi)->max_nid *
NM_I(sbi)->dirty_nats_ratio / 100;
}
static inline bool excess_cached_nats(struct f2fs_sb_info *sbi)
{
- return NM_I(sbi)->nat_cnt >= DEF_NAT_CACHE_THRESHOLD;
+ return NM_I(sbi)->nat_cnt[TOTAL_NAT] >= DEF_NAT_CACHE_THRESHOLD;
}
static inline bool excess_dirty_nodes(struct f2fs_sb_info *sbi)
diff --git a/fs/f2fs/shrinker.c b/fs/f2fs/shrinker.c
index d66de5999a26d..dd3c3c7a90ec8 100644
--- a/fs/f2fs/shrinker.c
+++ b/fs/f2fs/shrinker.c
@@ -18,9 +18,7 @@ static unsigned int shrinker_run_no;
static unsigned long __count_nat_entries(struct f2fs_sb_info *sbi)
{
- long count = NM_I(sbi)->nat_cnt - NM_I(sbi)->dirty_nat_cnt;
-
- return count > 0 ? count : 0;
+ return NM_I(sbi)->nat_cnt[RECLAIMABLE_NAT];
}
static unsigned long __count_free_nids(struct f2fs_sb_info *sbi)
--
2.27.0
next prev parent reply other threads:[~2021-01-04 16:05 UTC|newest]
Thread overview: 75+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-01-04 15:56 [PATCH 5.10 00/63] 5.10.5-rc1 review Greg Kroah-Hartman
2021-01-04 15:56 ` [PATCH 5.10 01/63] net/sched: sch_taprio: reset child qdiscs before freeing them Greg Kroah-Hartman
2021-01-04 22:58 ` Sasha Levin
2021-01-04 23:06 ` Jakub Kicinski
2021-01-04 15:56 ` [PATCH 5.10 02/63] mptcp: fix security context on server socket Greg Kroah-Hartman
2021-01-04 15:56 ` [PATCH 5.10 03/63] ethtool: fix error paths in ethnl_set_channels() Greg Kroah-Hartman
2021-01-04 15:56 ` [PATCH 5.10 04/63] ethtool: fix string set id check Greg Kroah-Hartman
2021-01-04 15:56 ` [PATCH 5.10 05/63] md/raid10: initialize r10_bio->read_slot before use Greg Kroah-Hartman
2021-01-04 15:56 ` [PATCH 5.10 06/63] drm/amd/display: Add get_dig_frontend implementation for DCEx Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 07/63] io_uring: close a small race gap for files cancel Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 08/63] jffs2: Allow setting rp_size to zero during remounting Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 09/63] jffs2: Fix NULL pointer dereference in rp_size fs option parsing Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 10/63] spi: dw-bt1: Fix undefined devm_mux_control_get symbol Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 11/63] opp: fix memory leak in _allocate_opp_table Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 12/63] opp: Call the missing clk_put() on error Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 13/63] scsi: block: Fix a race in the runtime power management code Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 14/63] mm/hugetlb: fix deadlock in hugetlb_cow error path Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 15/63] mm: memmap defer init doesnt work as expected Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 16/63] lib/zlib: fix inflating zlib streams on s390 Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 17/63] io_uring: dont assume mm is constant across submits Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 18/63] io_uring: use bottom half safe lock for fixed file data Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 19/63] io_uring: add a helper for setting a ref node Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 20/63] io_uring: fix io_sqe_files_unregister() hangs Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 21/63] kernel/io_uring: cancel io_uring before task works Greg Kroah-Hartman
2021-01-04 16:06 ` Pavel Begunkov
2021-01-04 17:43 ` Sasha Levin
2021-01-04 15:57 ` [PATCH 5.10 22/63] uapi: move constants from <linux/kernel.h> to <linux/const.h> Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 23/63] tools headers UAPI: Sync linux/const.h with the kernel headers Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 24/63] cgroup: Fix memory leak when parsing multiple source parameters Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 25/63] zlib: move EXPORT_SYMBOL() and MODULE_LICENSE() out of dfltcc_syms.c Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 26/63] scsi: cxgb4i: Fix TLS dependency Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 27/63] Bluetooth: hci_h5: close serdev device and free hu in h5_close Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 28/63] fbcon: Disable accelerated scrolling Greg Kroah-Hartman
2021-01-07 8:13 ` Geert Uytterhoeven
2021-01-04 15:57 ` [PATCH 5.10 29/63] reiserfs: add check for an invalid ih_entry_count Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 30/63] misc: vmw_vmci: fix kernel info-leak by initializing dbells in vmci_ctx_get_chkpt_doorbells() Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 31/63] media: gp8psk: initialize stats at power control logic Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 32/63] f2fs: fix shift-out-of-bounds in sanity_check_raw_super() Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 33/63] ALSA: seq: Use bool for snd_seq_queue internal flags Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 34/63] ALSA: rawmidi: Access runtime->avail always in spinlock Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 35/63] bfs: dont use WARNING: string when its just info Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 36/63] ext4: check for invalid block size early when mounting a file system Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 37/63] fcntl: Fix potential deadlock in send_sig{io, urg}() Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 38/63] io_uring: check kthread stopped flag when sq thread is unparked Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 39/63] rtc: sun6i: Fix memleak in sun6i_rtc_clk_init Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 40/63] module: set MODULE_STATE_GOING state when a module fails to load Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 41/63] quota: Dont overflow quota file offsets Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 42/63] rtc: pl031: fix resource leak in pl031_probe Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 43/63] powerpc: sysdev: add missing iounmap() on error in mpic_msgr_probe() Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 44/63] i3c master: fix missing destroy_workqueue() on error in i3c_master_register Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 45/63] NFSv4: Fix a pNFS layout related use-after-free race when freeing the inode Greg Kroah-Hartman
2021-01-04 15:57 ` Greg Kroah-Hartman [this message]
2021-01-04 15:57 ` [PATCH 5.10 47/63] f2fs: fix race of pending_pages in decompression Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 48/63] module: delay kobject uevent until after module init call Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 49/63] powerpc/64: irq replay remove decrementer overflow check Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 50/63] fs/namespace.c: WARN if mnt_count has become negative Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 51/63] watchdog: rti-wdt: fix reference leak in rti_wdt_probe Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 52/63] um: random: Register random as hwrng-core device Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 53/63] um: ubd: Submit all data segments atomically Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 54/63] NFSv4.2: Dont error when exiting early on a READ_PLUS buffer overflow Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 55/63] ceph: fix inode refcount leak when ceph_fill_inode on non-I_NEW inode fails Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 56/63] drm/amd/display: updated wm table for Renoir Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 57/63] tick/sched: Remove bogus boot "safety" check Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 58/63] s390: always clear kernel stack backchain before calling functions Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 59/63] io_uring: remove racy overflow list fast checks Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 60/63] ALSA: pcm: Clear the full allocated memory at hw_params Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 61/63] dm verity: skip verity work if I/O error when system is shutting down Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 62/63] ext4: avoid s_mb_prefetch to be zero in individual scenarios Greg Kroah-Hartman
2021-01-04 15:57 ` [PATCH 5.10 63/63] device-dax: Fix range release Greg Kroah-Hartman
2021-01-05 6:06 ` [PATCH 5.10 00/63] 5.10.5-rc1 review Daniel Díaz
2021-01-05 12:55 ` Jeffrin Jose T
2021-01-05 13:05 ` Greg Kroah-Hartman
2021-01-06 18:43 ` Jeffrin Jose T
2021-01-05 16:38 ` Shuah Khan
2021-01-05 18:17 ` Guenter Roeck
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210104155711.049210727@linuxfoundation.org \
--to=gregkh@linuxfoundation.org \
--cc=Light.Hsieh@mediatek.com \
--cc=jaegeuk@kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=sashal@kernel.org \
--cc=stable@vger.kernel.org \
--cc=yuchao0@huawei.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).