f2fs: fix long latency due to discard during umount
PATCH v3 1/1 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. In that case, submit_bio() will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.
Fix this by submitting the discard requests with REQ_NOWAIT flags during umount. This will return -EAGAIN for UFS queue/tag full scenario without waiting in the context of submit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.
---
v3:
-Handle the regression reported by Chao with v2.
-simplify the logic to split the dc with multiple bios incase any bio returns
EAGAIN and retry those new dc within 5 sec timeout.
fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)
fs/f2fs/segment.c +51 -14
@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);
@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;
@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);
@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);
@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }
@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))
@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;
@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;
Signature · 2 lines
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.
kbuild test robot
build botThank you for the patch! Yet something to improve:
Hi Sahitya,
Thank you for the patch! Yet something to improve:
[auto build test ERROR on v5.6]
[cannot apply to f2fs/dev-test next-20200327]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
url: https://github.com/0day-ci/linux/commits/Sahitya-Tummala/f2fs-fix-long-latency-due-to-discard-during-umount/20200330-151252 base: 7111951b8d4973bda27ff663f2cf18b663d15b48 config: nios2-randconfig-a001-20200329 (attached as .config) compiler: nios2-linux-gcc (GCC) 9.3.0
reproduce:
wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross chmod +x ~/bin/make.cross
# save the attached .config to linux build tree
GCC_VERSION=9.3.0 make.cross ARCH=nios2
If you fix the issue, kindly add following tag
All errors (new ones prefixed by >>):
fs/f2fs/segment.c: In function '__issue_discard_cmd':
>> fs/f2fs/segment.c:1542:7: error: 'DEFAULT_IO_TIMEOUT' undeclared (first use in this function); did you mean 'BLK_DEFAULT_SG_TIMEOUT'?
1542 | DEFAULT_IO_TIMEOUT);
| ^~~~~~~~~~~~~~~~~~
| BLK_DEFAULT_SG_TIMEOUT fs/f2fs/segment.c:1542:7: note: each undeclared identifier is reported only once for each function it appears in
vim +1542 fs/f2fs/segment.c
1489
1490 static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
1491 struct discard_policy *dpolicy)
1492 {
1493 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1494 struct list_head *pend_list;
1495 struct discard_cmd *dc, *tmp;
1496 struct blk_plug plug;
1497 int i, err, issued = 0;
1498 bool io_interrupted = false;
1499 bool retry;
1500
1501 if (dpolicy->timeout != 0)
1502 f2fs_update_time(sbi, dpolicy->timeout);
1503
1504 retry:
1505 retry = false;
1506 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1507 if (dpolicy->timeout != 0 &&
1508 f2fs_time_over(sbi, dpolicy->timeout))
1509 break;
1510
1511 if (i + 1 < dpolicy->granularity)
1512 break;
1513
1514 if (i < DEFAULT_DISCARD_GRANULARITY && dpolicy->ordered)
1515 return __issue_discard_cmd_orderly(sbi, dpolicy);
1516
1517 pend_list = &dcc->pend_list[i];
1518
1519 mutex_lock(&dcc->cmd_lock);
1520 if (list_empty(pend_list))
1521 goto next;
1522 if (unlikely(dcc->rbtree_check))
1523 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
1524 &dcc->root));
1525 blk_start_plug(&plug);
1526 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1527 f2fs_bug_on(sbi, dc->state != D_PREP);
1528
1529 if (dpolicy->timeout != 0 &&
1530 f2fs_time_over(sbi, dpolicy->timeout))
1531 break;
1532
1533 if (dpolicy->io_aware && i < dpolicy->io_aware_gran &&
1534 !is_idle(sbi, DISCARD_TIME)) {
1535 io_interrupted = true;
1536 break;
1537 }
1538
1539 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
1540 if (err == -EAGAIN) {
1541 congestion_wait(BLK_RW_ASYNC,
> 1542 DEFAULT_IO_TIMEOUT);
1543 retry = true;
1544 }
1545
1546 if (issued >= dpolicy->max_requests)
1547 break;
1548 }
1549 blk_finish_plug(&plug);
1550 next:
1551 mutex_unlock(&dcc->cmd_lock);
1552
1553 if (issued >= dpolicy->max_requests || io_interrupted)
1554 break;
1555 }
1556
1557 if (retry)
1558 goto retry;
1559
1560 if (!issued && io_interrupted)
1561 issued = -1;
1562
1563 return issued;
1564 }
1565
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
kbuild test robot
build botThank you for the patch! Yet something to improve:
Hi Sahitya,
Thank you for the patch! Yet something to improve:
[auto build test ERROR on v5.6]
[cannot apply to f2fs/dev-test next-20200327]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system. BTW, we also suggest to use '--base' option to specify the base tree in git format-patch, please see https://stackoverflow.com/a/37406982]
url: https://github.com/0day-ci/linux/commits/Sahitya-Tummala/f2fs-fix-long-latency-due-to-discard-during-umount/20200330-151252 base: 7111951b8d4973bda27ff663f2cf18b663d15b48 config: x86_64-randconfig-s0-20200330 (attached as .config) compiler: gcc-5 (Ubuntu 5.5.0-12ubuntu1) 5.5.0 20171010 reproduce: # save the attached .config to linux build tree make ARCH=x86_64
If you fix the issue, kindly add following tag
All errors (new ones prefixed by >>):
fs/f2fs/segment.c: In function '__issue_discard_cmd':
>> fs/f2fs/segment.c:1542:7: error: 'DEFAULT_IO_TIMEOUT' undeclared (first use in this function)
DEFAULT_IO_TIMEOUT);
^ fs/f2fs/segment.c:1542:7: note: each undeclared identifier is reported only once for each function it appears in
vim +/DEFAULT_IO_TIMEOUT +1542 fs/f2fs/segment.c
1489
1490 static int __issue_discard_cmd(struct f2fs_sb_info *sbi,
1491 struct discard_policy *dpolicy)
1492 {
1493 struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
1494 struct list_head *pend_list;
1495 struct discard_cmd *dc, *tmp;
1496 struct blk_plug plug;
1497 int i, err, issued = 0;
1498 bool io_interrupted = false;
1499 bool retry;
1500
1501 if (dpolicy->timeout != 0)
1502 f2fs_update_time(sbi, dpolicy->timeout);
1503
1504 retry:
1505 retry = false;
1506 for (i = MAX_PLIST_NUM - 1; i >= 0; i--) {
1507 if (dpolicy->timeout != 0 &&
1508 f2fs_time_over(sbi, dpolicy->timeout))
1509 break;
1510
1511 if (i + 1 < dpolicy->granularity)
1512 break;
1513
1514 if (i < DEFAULT_DISCARD_GRANULARITY && dpolicy->ordered)
1515 return __issue_discard_cmd_orderly(sbi, dpolicy);
1516
1517 pend_list = &dcc->pend_list[i];
1518
1519 mutex_lock(&dcc->cmd_lock);
1520 if (list_empty(pend_list))
1521 goto next;
1522 if (unlikely(dcc->rbtree_check))
1523 f2fs_bug_on(sbi, !f2fs_check_rb_tree_consistence(sbi,
1524 &dcc->root));
1525 blk_start_plug(&plug);
1526 list_for_each_entry_safe(dc, tmp, pend_list, list) {
1527 f2fs_bug_on(sbi, dc->state != D_PREP);
1528
1529 if (dpolicy->timeout != 0 &&
1530 f2fs_time_over(sbi, dpolicy->timeout))
1531 break;
1532
1533 if (dpolicy->io_aware && i < dpolicy->io_aware_gran &&
1534 !is_idle(sbi, DISCARD_TIME)) {
1535 io_interrupted = true;
1536 break;
1537 }
1538
1539 err = __submit_discard_cmd(sbi, dpolicy, dc, &issued);
1540 if (err == -EAGAIN) {
1541 congestion_wait(BLK_RW_ASYNC,
> 1542 DEFAULT_IO_TIMEOUT);
1543 retry = true;
1544 }
1545
1546 if (issued >= dpolicy->max_requests)
1547 break;
1548 }
1549 blk_finish_plug(&plug);
1550 next:
1551 mutex_unlock(&dcc->cmd_lock);
1552
1553 if (issued >= dpolicy->max_requests || io_interrupted)
1554 break;
1555 }
1556
1557 if (retry)
1558 goto retry;
1559
1560 if (!issued && io_interrupted)
1561 issued = -1;
1562
1563 return issued;
1564 }
1565
---
0-DAY CI Kernel Test Service, Intel Corporation
https://lists.01.org/hyperkitty/list/kbuild-all(a)lists.01.org
Jaegeuk Kim
maintainerQuoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.
Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in __issue_discard_cmd()?
Quoted 149 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.
Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?
Do you mean something like this?
fs/f2fs/segment.c +9 -2
@@ -1099,7 +1099,7 @@ 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->max_requests =30 ; dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;
@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))
@@ -1522,6 +1524,11 @@ 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 ;
Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.
Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Quoted 27 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
fs/f2fs/segment.c +0
-0
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 5 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;
8 or 16?
It looks more simple than previous implementation.
Thanks,
Quoted 180 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 147 lines
Quoted 147 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;
Quoted 4 lines
Quoted 3 lines
Quoted 3 lines
Signature · 2 lines
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.
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
Jaegeuk Kim
maintainerQuoted 27 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
fs/f2fs/segment.c +0
-0
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 5 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;
Can we use max queue depth of the block device?
Quoted 188 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Hi Jaegeuk,
Quoted 23 lines · Jaegeuk Kim
Quoted 22 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
Quoted 12 lines
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 10 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;Can we use max queue depth of the block device?
I think it should be limited to 8 or 16 as Chao suggested, so that we can have better control on the given timeout value? Thoughts?
Thanks,
Quoted 189 lines
Quoted 189 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Jaegeuk Kim
maintainerQuoted 39 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 23 lines · Jaegeuk Kim
Quoted 22 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
Quoted 12 lines
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 10 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;Can we use max queue depth of the block device?
Quoted 2 lines
I think it should be limited to 8 or 16 as Chao suggested, so that we can have better control on the given timeout value? Thoughts?
Where is 8 or 16 coming from? What about SSD? Sorry, it's unclear to me.
Quoted 197 lines
Thanks,
Quoted 189 lines
Quoted 189 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Quoted 40 lines · Jaegeuk Kim
Quoted 39 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 35 lines · Jaegeuk Kim
Quoted 27 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
fs/f2fs/segment.c +0
-0fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 5 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;Can we use max queue depth of the block device?
Quoted 4 lines
Quoted 2 lines
I think it should be limited to 8 or 16 as Chao suggested, so that we can have better control on the given timeout value? Thoughts?
Where is 8 or 16 coming from? What about SSD? Sorry, it's unclear to me.
With this patch we now wait for a batch of discard requests in __issue_discard_cmd() with a upper timeout of 5 sec. So, I thought that having a smaller batch of discard requests would help us to avoid queuing more requests and end up waiting for more time in __wait_all_discard_cmd(). Today we have DEF_MAX_DISCARD_REQUEST as 8 for default max discard requests. If it too less, may be for umount we can use 16 instead. But the idea is to avoid more wait time.
If you would like to propose to use max queue depth, we can get it from bdev as sbi->sb->s_bdev->bd_queue->queue_depth. Is it okay to use it in our FS? or you think it should be hardcoded to 32 (as most UFS/eMMC devices will have this as its queue depth)? Please let me know.
Thanks,
Quoted 198 lines
Quoted 198 lines
Thanks,
Quoted 189 lines
Quoted 189 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Jaegeuk Kim
maintainerQuoted 44 lines · Sahitya Tummala
Quoted 43 lines · Jaegeuk Kim
Quoted 41 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 23 lines · Jaegeuk Kim
Quoted 22 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
Quoted 12 lines
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 10 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;Can we use max queue depth of the block device?
I think it should be limited to 8 or 16 as Chao suggested, so that we can have better control on the given timeout value? Thoughts?
Quoted 8 lines
Quoted 1 line
Where is 8 or 16 coming from? What about SSD? Sorry, it's unclear to me.
With this patch we now wait for a batch of discard requests in
__issue_discard_cmd()with a upper timeout of 5 sec. So, I thought that having a smaller batch of discard requests would help us to avoid queuing more requests and end up waiting for more time in__wait_all_discard_cmd(). Today we haveDEF_MAX_DISCARD_REQUESTas 8 for default max discard requests. If it too less, may be for umount we can use 16 instead. But the idea is to avoid more wait time.
Ok, then, it seems we can just remove this line to use dpolicy->max_requests, and retry issuing all the discard commands like below. Later, I thnk it'd be fine to tune the max_requests based on bd_queue->queue_depth.
Quoted 202 lines
If you would like to propose to use max queue depth, we can get it from bdev as
sbi->sb->s_bdev->bd_queue->queue_depth. Is it okay to use it in our FS? or you think it should be hardcoded to 32 (as most UFS/eMMC devices will have this as its queue depth)? Please let me know.Thanks,
Quoted 194 lines
Quoted 194 lines
Thanks,
Quoted 189 lines
Quoted 189 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
Jaegeuk Kim
maintainerHi Sahitya,
Could you please post the revised patch?
Thanks,
Quoted 271 lines · Jaegeuk Kim
Quoted 44 lines · Sahitya Tummala
Quoted 43 lines · Jaegeuk Kim
Quoted 41 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 23 lines · Jaegeuk Kim
Quoted 22 lines · Sahitya Tummala
Hi Jaegeuk,
Quoted 8 lines · Jaegeuk Kim
Quoted 7 lines · Sahitya Tummala
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. In that case,
submit_bio()will wait for the already queued discard requests to complete to get a free tag, which can potentially take way more than 5 sec.Quoted 8 lines
Quoted 5 lines
Fix this by submitting the discard requests with
REQ_NOWAITflags during umount. This will return-EAGAINfor UFS queue/tag full scenario without waiting in the context ofsubmit_bio(). The FS can then handle these requests by retrying again within the stipulated discard timeout period to avoid long latencies.Sorry, Sahitya, but, do we really need to do like this? How about just controlling # of outstanding discarding bios in
__issue_discard_cmd()?Do you mean something like this?
Quoted 12 lines
fs/f2fs/segment.c static void __init_discard_policy(struct f2fs_sb_info *sbi, · line 1102
Quoted 10 lines
fs/f2fs/segment.c +1
-1@@ -1099,7 +1099,7 @@ 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->max_requests =30 ;Can we use max queue depth of the block device?
I think it should be limited to 8 or 16 as Chao suggested, so that we can have better control on the given timeout value? Thoughts?
Quoted 8 lines
Quoted 1 line
Where is 8 or 16 coming from? What about SSD? Sorry, it's unclear to me.
With this patch we now wait for a batch of discard requests in
__issue_discard_cmd()with a upper timeout of 5 sec. So, I thought that having a smaller batch of discard requests would help us to avoid queuing more requests and end up waiting for more time in__wait_all_discard_cmd(). Today we haveDEF_MAX_DISCARD_REQUESTas 8 for default max discard requests. If it too less, may be for umount we can use 16 instead. But the idea is to avoid more wait time.Ok, then, it seems we can just remove this line to use
dpolicy->max_requests, and retry issuing all the discard commands like below. Later, I thnk it'd be fine to tune themax_requestsbased onbd_queue->queue_depth.Quoted 202 lines
If you would like to propose to use max queue depth, we can get it from bdev as
sbi->sb->s_bdev->bd_queue->queue_depth. Is it okay to use it in our FS? or you think it should be hardcoded to 32 (as most UFS/eMMC devices will have this as its queue depth)? Please let me know.Thanks,
Quoted 194 lines
Quoted 194 lines
Thanks,
Quoted 189 lines
Quoted 189 lines
fs/f2fs/segment.c +8
-1dpolicy->io_aware = false;/* we need to issue all to keep CP_TRIMMED_FLAG */ dpolicy->granularity =1 ;@@ -1470,12 +1470,14 @@ 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 !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + issued =0 ;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1522,6 +1524,11 @@ 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 ;Thanks,
Quoted 150 lines
Quoted 150 lines
Signed-off-by Sahitya Tummala <[email protected]>---
v3:(quoted diff) +0
-2-Handle the regression reported by Chao with v2. -simplify the logic to split the dc with multiple bios incase any bio returns EAGAIN and retry those new dc within 5 sec timeout. fs/f2fs/segment.c | 65 +++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 51 insertions(+), 14 deletions(-)fs/f2fs/segment.c +51
-14@@ -1029,13 +1029,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;unsigned long flags; - dc->error =blk_status_to_errno (bio->bi_status); -spin_lock_irqsave (&dc->lock, flags); +if (!dc->error) + dc->error =blk_status_to_errno (bio->bi_status); + dc->bio_ref--; -if (!dc->bio_ref && dc->state ==D_SUBMIT ) { - dc->state =D_DONE ; -complete_all (&dc->wait); +if (!dc->bio_ref) { +if (dc->error || dc->state ==D_SUBMIT ) { + dc->state =D_DONE ; +complete_all (&dc->wait); + } }spin_unlock_irqrestore (&dc->lock, flags);bio_put (bio);@@ -1124,10 +1127,13 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,struct discard_cmd_control *dcc =SM_I (sbi)->dcc_info;struct list_head *wait_list = (dpolicy->type ==DPOLICY_FSTRIM ) ? &(dcc->fstrim_list) : &(dcc->wait_list); -int flag = dpolicy->sync ?REQ_SYNC :0 ; +int flag;block_t lstart, start, len, total_len;int err =0 ; + flag = dpolicy->sync ?REQ_SYNC :0 ; + flag |= dpolicy->type ==DPOLICY_UMOUNT ?REQ_NOWAIT :0 ; +if (dc->state !=D_PREP )return 0 ;@@ -1192,10 +1198,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, dc->bio_ref++;spin_unlock_irqrestore (&dc->lock, flags); -atomic_inc (&dcc->queued_discard); - dc->queued++; -list_move_tail (&dc->list, wait_list); -/* sanity check on discard range */ __check_sit_bitmap (sbi, lstart, lstart + len);@@ -1203,6 +1205,29 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, bio->bi_end_io = f2fs_submit_discard_endio; bio->bi_opf |= flag;submit_bio (bio); +if (flag &REQ_NOWAIT ) { +if (dc->error == -EAGAIN ) { +spin_lock_irqsave (&dc->lock, flags); + dc->len -= len; +if (!dc->len) { + dc->len = total_len; + dc->state =D_PREP ; +reinit_completion (&dc->wait); + }else { + dcc->undiscard_blks -= total_len; +if (dc->state ==D_PARTIAL ) + dc->state =D_SUBMIT ; + } + err = dc->error; + dc->error =0 ; +spin_unlock_irqrestore (&dc->lock, flags); +break ; + } + } + +atomic_inc (&dcc->queued_discard); + dc->queued++; +list_move_tail (&dc->list, wait_list);atomic_inc (&dcc->issued_discard);@@ -1214,8 +1239,9 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi, len = total_len; } -if (!err && len) -__update_discard_tree_range (sbi, bdev, lstart, start, len); +if ((!err || err == -EAGAIN ) && total_len && dc->start != start) +__update_discard_tree_range (sbi, bdev, lstart, start, + total_len);return err; }@@ -1470,12 +1496,15 @@ 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, err, issued =0 ;bool io_interrupted = false; +bool retry;if (dpolicy->timeout !=0 )f2fs_update_time (sbi, dpolicy->timeout); +retry: + retry = false;for (i =MAX_PLIST_NUM -1 ; i >=0 ; i--) {if (dpolicy->timeout !=0 &&f2fs_time_over (sbi, dpolicy->timeout))@@ -1509,7 +1538,12 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } -__submit_discard_cmd (sbi, dpolicy, dc, &issued); + err =__submit_discard_cmd (sbi, dpolicy, dc, &issued); +if (err == -EAGAIN ) { +congestion_wait (BLK_RW_ASYNC , +DEFAULT_IO_TIMEOUT ); + retry = true; + }if (issued >= dpolicy->max_requests)break ;@@ -1522,6 +1556,9 @@ static int __issue_discard_cmd(struct f2fs_sb_info *sbi,break ; } +if (retry) + goto retry; +if (!issued && io_interrupted) issued =-1 ;Signature · 2 lines
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.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.Quoted 9 lines
Quoted 4 lines
Quoted 4 lines
Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.Signature · 3 lines
-- Sent by a consultant of the Qualcomm Innovation Center, Inc. The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum.
Quoted 4 lines
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
_______________________________________________
Linux-f2fs-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel