On 2018/12/19 6:43, Jaegeuk Kim wrote: > On 12/18, Chao Yu wrote: >> On 2018/12/14 13:01, Jaegeuk Kim wrote: >>> Sometimes, I could observe # of issuing_discard to be 1 which blocks >>> background >>> jobs due to is_idle()=false. >>> The only way to get out of it was to trigger gc_urgent. This patch avoids >>> that >>> by checking any candidates as done in the list. >> >> Well, as below code, once we issued discard commands, we will wait all >> queued discard end their IO, so do you know what flow can cause such >> condition...? > > It's very subtle, and I suspect somehow race condition not from here. > >> >> issued = __issue_discard_cmd(sbi, &dpolicy); >> if (issued > 0) { >> __wait_all_discard_cmd(sbi, &dpolicy); >> >> Or, I doubt that 'issued' statistical info could be wrong. > > No, it simply came back to 0 with this patch. So, something is pending > in the queue even if it is done.
__submit_discard_cmd() { .... dc->len = 0; <---- reset discard length here while (total_len && *issued < dpolicy->max_requests && !err) { struct bio *bio = NULL; unsigned long flags; bool last = true; if (len > max_discard_blocks) { len = max_discard_blocks; <---- will truncate len to max_discard_blocks last = false; } (*issued)++; if (*issued == dpolicy->max_requests) last = true; dc->len += len; <----- will update len into dc->len ... } __wait_discard_cmd_range() { ... if (dc->len < dpolicy->granularity) continue; So we can only remove discard entry which size is not smaller than dpolicy->granularity, if max_discard_blocks is smaller than granularity, after splitting discard in __submit_discard_cmd(), we may left small-sized entry there. ... } Could you check that whether the left discard was skipped being removed due to above reason? Thanks > > Thanks, > >> >> Thanks, >> >>> >>> Signed-off-by: Jaegeuk Kim <jaeg...@kernel.org> >>> --- >>> fs/f2fs/segment.c | 4 ++++ >>> 1 file changed, 4 insertions(+) >>> >>> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c >>> index 49ea9009ab5a..acbbc924e518 100644 >>> --- a/fs/f2fs/segment.c >>> +++ b/fs/f2fs/segment.c >>> @@ -1651,6 +1651,10 @@ static int issue_discard_thread(void *data) >>> if (dcc->discard_wake) >>> dcc->discard_wake = 0; >>> >>> + /* clean up pending candidates before going to sleep */ >>> + if (atomic_read(&dcc->queued_discard)) >>> + __wait_all_discard_cmd(sbi, NULL); >>> + >>> if (try_to_freeze()) >>> continue; >>> if (f2fs_readonly(sbi->sb)) >>> > > . >