2020-04-15 22:44:38

by Sahitya Tummala

Subject: [PATCH v5] f2fs: fix long latency due to discard during umount

F2FS already has a default timeout of 5 secs for discards that
can be issued during umount, but it can take more than the 5 sec
timeout if the underlying UFS device queue is already full and there
are no more available free tags to be used. Fix this by submitting a
small batch of discard requests so that it won't cause the device
queue to be full at any time and thus doesn't incur its wait time
in the umount context.

Signed-off-by: Sahitya Tummala <[email protected]>
---
v5:
- rebase on Jaegeuk's dev branch.
- add a missing change to fix compilation issue.

fs/f2fs/segment.c | 12 ++++++++++--
1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index b7a9421..90c7582 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -1101,7 +1101,6 @@ static void __init_discard_policy(struct f2fs_sb_info *sbi,
} else if (discard_type == DPOLICY_FSTRIM) {
dpolicy->io_aware = false;
} else if (discard_type == DPOLICY_UMOUNT) {
- dpolicy->max_requests = UINT_MAX;
dpolicy->io_aware = false;
/* we need to issue all to keep CP_TRIMMED_FLAG */
dpolicy->granularity = 1;
@@ -1463,6 +1462,8 @@ static unsigned int __issue_discard_cmd_orderly(struct f2fs_sb_info *sbi,

return issued;
}
+static unsigned int __wait_all_discard_cmd(struct f2fs_sb_info *sbi,
+ struct discard_policy *dpolicy);

static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
struct discard_policy *dpolicy)
@@ -1471,12 +1472,14 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
struct list_head *pend_list;
struct discard_cmd *dc, *tmp;
struct blk_plug plug;
- int i, issued = 0;
+ int i, issued;
bool io_interrupted = false;

if (dpolicy->timeout)
f2fs_update_time(sbi, UMOUNT_DISCARD_TIMEOUT);

+retry:
+ issued = 0;
for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
if (dpolicy->timeout &&
f2fs_time_over(sbi, UMOUNT_DISCARD_TIMEOUT))
@@ -1523,6 +1526,11 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
break;
}

+ if (dpolicy->type == DPOLICY_UMOUNT && issued) {
+ __wait_all_discard_cmd(sbi, dpolicy);
+ goto retry;
+ }
+
if (!issued && io_interrupted)
issued = -1;

--
Qualcomm India Private Limited, on behalf of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project.

2020-04-16 06:12:46

by Chao Yu

Subject: Re: [PATCH v5] f2fs: fix long latency due to discard during umount

On 2020/4/15 17:07, Sahitya Tummala wrote:
> F2FS already has a default timeout of 5 secs for discards that
> can be issued during umount, but it can take more than the 5 sec
> timeout if the underlying UFS device queue is already full and there
> are no more available free tags to be used. Fix this by submitting a
> small batch of discard requests so that it won't cause the device
> queue to be full at any time and thus doesn't incur its wait time
> in the umount context.
>
> Signed-off-by: Sahitya Tummala <[email protected]>

Reviewed-by: Chao Yu <[email protected]>

Thanks,