Skip to contentSkip to conversation

f2fs: add sysfs entry to avoid FUA

PATCH In review Jaegeuk Kim 5 participants 10 messages May 27, 2022 – May 31, 2022

Subsystems (2)
lore
Share & tools
lore

Jaegeuk Kim

author maintainer
#0 lore raw

PATCH 1/1 f2fs: add sysfs entry to avoid FUA

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

---
 Documentation/ABI/testing/sysfs-fs-f2fs | 7 +++++++
 fs/f2fs/data.c                          | 2 ++
 fs/f2fs/f2fs.h                          | 1 +
 fs/f2fs/sysfs.c                         | 2 ++
 4 files changed, 12 insertions(+)
Documentation/ABI/testing/sysfs-fs-f2fs +7 -0
index 9b583dd0298b..cd96b09d7182 100644
@@ -434,6 +434,7 @@ Date:		April 2020
 Contact:	"Daeho Jeong" <[email protected]>
 Description:	Give a way to change iostat_period time. 3secs by default.
 		The new iostat trace gives stats gap given the period.
+
 What:		/sys/fs/f2fs/<disk>/max_io_bytes
 Date:		December 2020
 Contact:	"Jaegeuk Kim" <[email protected]>
@@ -442,6 +443,12 @@ Description:	This gives a control to limit the bio size in f2fs.
 		whereas, if it has a certain bytes value, f2fs won't submit a
 		bio larger than that size.
 
+What:		/sys/fs/f2fs/<disk>/no_fua_dio
+Date:		May 2022
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	This gives a signal to iomap, which should not use FUA for
+		direct IOs. Default: 0.
+
 What:		/sys/fs/f2fs/<disk>/stat/sb_status
 Date:		December 2020
 Contact:	"Chao Yu" <[email protected]>
fs/f2fs/data.c +2 -0
index f5f2b7233982..23486486eab2 100644
@@ -4153,6 +4153,8 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
 	    offset + length > i_size_read(inode))
 		iomap->flags |= IOMAP_F_DIRTY;
+	if (F2FS_I_SB(inode)->no_fua_dio)
+		iomap->flags |= IOMAP_F_DIRTY;
 
 	return 0;
 }
fs/f2fs/f2fs.h +1 -0
index e10838879538..c2400ea0080b 100644
@@ -1671,6 +1671,7 @@ struct f2fs_sb_info {
 	int dir_level;				/* directory level */
 	int readdir_ra;				/* readahead inode in readdir */
 	u64 max_io_bytes;			/* max io bytes to merge IOs */
+	int no_fua_dio;				/* avoid FUA in DIO */
 
 	block_t user_block_count;		/* # of user blocks */
 	block_t total_valid_block_count;	/* # of valid blocks */
fs/f2fs/sysfs.c +2 -0
index 4c50aedd5144..24d628ca92cc 100644
@@ -771,6 +771,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, iostat_period_ms, iostat_period_ms);
 #endif
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, readdir_ra, readdir_ra);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_io_bytes, max_io_bytes);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, no_fua_dio, no_fua_dio);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_super_block, extension_list, extension_list);
 #ifdef CONFIG_F2FS_FAULT_INJECTION
@@ -890,6 +891,7 @@ static struct attribute *f2fs_attrs[] = {
 #endif
 	ATTR_LIST(readdir_ra),
 	ATTR_LIST(max_io_bytes),
+	ATTR_LIST(no_fua_dio),
 	ATTR_LIST(gc_pin_file_thresh),
 	ATTR_LIST(extension_list),
 #ifdef CONFIG_F2FS_FAULT_INJECTION
Signature · 1 line
2.36.1.124.g0e6072fb45-goog
#1 lore raw

[+Cc linux-block for FUA, and linux-xfs for iomap]

Quoted 4 lines · Jaegeuk Kim

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

Quoted 20 lines
---
 Documentation/ABI/testing/sysfs-fs-f2fs | 7 +++++++
 fs/f2fs/data.c                          | 2 ++
 fs/f2fs/f2fs.h                          | 1 +
 fs/f2fs/sysfs.c                         | 2 ++
 4 files changed, 12 insertions(+)
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
index 9b583dd0298b..cd96b09d7182 100644
@@ -434,6 +434,7 @@ Date:		April 2020
 Contact:	"Daeho Jeong" <[email protected]>
 Description:	Give a way to change iostat_period time. 3secs by default.
 		The new iostat trace gives stats gap given the period.
+
 What:		/sys/fs/f2fs/<disk>/max_io_bytes
 Date:		December 2020
 Contact:	"Jaegeuk Kim" <[email protected]>

Documentation/ABI/testing/sysfs-fs-f2fs Description: This gives a control to limit the bio size in f2fs. · line 450

Quoted 8 lines
Documentation/ABI/testing/sysfs-fs-f2fs +5 -0
@@ -442,6 +443,12 @@ Description:	This gives a control to limit the bio size in f2fs.
 		whereas, if it has a certain bytes value, f2fs won't submit a
 		bio larger than that size.
 
+What:		/sys/fs/f2fs/<disk>/no_fua_dio
+Date:		May 2022
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	This gives a signal to iomap, which should not use FUA for
+		direct IOs. Default: 0.

iomap is an implementation detail, so it shouldn't be mentioned in UAPI documentation. UAPI documentation should describe user-visible behavior only.

Quoted 9 lines
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
+
 What:		/sys/fs/f2fs/<disk>/stat/sb_status
 Date:		December 2020
 Contact:	"Chao Yu" <[email protected]>
fs/f2fs/data.c +0 -0
index f5f2b7233982..23486486eab2 100644

fs/f2fs/data.c static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length, · line 4157

Quoted 5 lines
fs/f2fs/data.c +2 -0
@@ -4153,6 +4153,8 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
 	    offset + length > i_size_read(inode))
 		iomap->flags |= IOMAP_F_DIRTY;
+	if (F2FS_I_SB(inode)->no_fua_dio)
+		iomap->flags |= IOMAP_F_DIRTY;

This is overloading the IOMAP_F_DIRTY flag to mean something other than dirty. Perhaps this flag needs to be renamed, or a new flag should be added?

fs/f2fs/f2fs.h struct f2fs_sb_info { · line 1674

Quoted 9 lines
fs/f2fs/f2fs.h +1 -0
index e10838879538..c2400ea0080b 100644
@@ -1671,6 +1671,7 @@ struct f2fs_sb_info {
 	int dir_level;				/* directory level */
 	int readdir_ra;				/* readahead inode in readdir */
 	u64 max_io_bytes;			/* max io bytes to merge IOs */
+	int no_fua_dio;				/* avoid FUA in DIO */

Make this a bool?

Quoted 13 lines
fs/f2fs/sysfs.c +1 -0
index 4c50aedd5144..24d628ca92cc 100644
@@ -771,6 +771,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, iostat_period_ms, iostat_period_ms);
 #endif
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, readdir_ra, readdir_ra);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_io_bytes, max_io_bytes);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, no_fua_dio, no_fua_dio);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_super_block, extension_list, extension_list);
 #ifdef CONFIG_F2FS_FAULT_INJECTION

fs/f2fs/sysfs.c static struct attribute *f2fs_attrs[] = { · line 894

Quoted 4 lines
fs/f2fs/sysfs.c +1 -0
@@ -890,6 +891,7 @@ static struct attribute *f2fs_attrs[] = {
 #endif
 	ATTR_LIST(readdir_ra),
 	ATTR_LIST(max_io_bytes),
+	ATTR_LIST(no_fua_dio),

Where is it validated that only valid values (0 or 1) can be written to this file?

- Eric

#2 lore raw
Quoted 1 line · Eric Biggers

[+Cc linux-block for FUA, and linux-xfs for iomap]

linux-fsdevel should really be used for iomap stuff...

Quoted 9 lines · Jaegeuk Kim
Quoted 4 lines

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

Agreed, this is a hardware problem that need to addressed with a driver quirk to stop it advertising FUA support. The high level fs/iomap code should always issue FUA writes where possible and the lower layers tell the block layer whether to issue the FUA as a FUA or write+cache flush pair.

And, quite frankly, exposing this sort of "hardware needs help" knob as a sysfs variable is exactly the sort of thing we should never do.

Users have no idea how to tune stuff like this correctly (even if they knew it existed!), yet we know exactly what hardware has this problem and the kernel already has mechanisms that would allow it to just Do The Right Thing. IOWs, we can fix this without the user even having to know that they have garbage hardware that needs special help....

Cheers,

Dave.

Signature · 2 lines
Dave Chinner
[email protected]
author maintainer
#3 lore raw
Quoted 12 lines · Dave Chinner
Quoted 1 line · Eric Biggers

[+Cc linux-block for FUA, and linux-xfs for iomap]

linux-fsdevel should really be used for iomap stuff...

Quoted 7 lines · Jaegeuk Kim
Quoted 4 lines

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Quoted 8 lines
Quoted 2 lines

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

Agreed, this is a hardware problem that need to addressed with a driver quirk to stop it advertising FUA support. The high level fs/iomap code should always issue FUA writes where possible and the lower layers tell the block layer whether to issue the FUA as a FUA or write+cache flush pair.

I was thinking to turn off FUA in driver side quickly tho, one concern was the bandwidth vs. latency. What if the device can support FUA having short latency while giving low bandwidth? In that case, we still have a room to utilize FUA for small-sized writes such as filesystem metadata writes, but avoid DIO w/ FUA for sequential write stream. Is this just HW problem? Or, does SW need to use FUA more efficiently?

Quoted 17 lines

And, quite frankly, exposing this sort of "hardware needs help" knob as a sysfs variable is exactly the sort of thing we should never do.

Users have no idea how to tune stuff like this correctly (even if they knew it existed!), yet we know exactly what hardware has this problem and the kernel already has mechanisms that would allow it to just Do The Right Thing. IOWs, we can fix this without the user even having to know that they have garbage hardware that needs special help....

Cheers,

Dave.

Signature · 2 lines
Dave Chinner
[email protected]
#4 lore raw
Quoted 22 lines · Jaegeuk Kim
Quoted 12 lines · Dave Chinner
Quoted 1 line · Eric Biggers

[+Cc linux-block for FUA, and linux-xfs for iomap]

linux-fsdevel should really be used for iomap stuff...

Quoted 7 lines · Jaegeuk Kim
Quoted 4 lines

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Quoted 8 lines
Quoted 2 lines

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

Agreed, this is a hardware problem that need to addressed with a driver quirk to stop it advertising FUA support. The high level fs/iomap code should always issue FUA writes where possible and the lower layers tell the block layer whether to issue the FUA as a FUA or write+cache flush pair.

Quoted 3 lines

I was thinking to turn off FUA in driver side quickly tho, one concern was the bandwidth vs. latency. What if the device can support FUA having short latency while giving low bandwidth?

Seriously, how is a user supposed to know this sort of thing about the hardware they are using? They don't, and to expect them to not only know about the existing of a weird sysfs knob, let alone how it applies to their hardware and their workload is totally unreasonable.

If the hardware has non-deterministic FUA write performance, or requires very careful switch over between cache flushes and FUA to get the most out of the hardware, then that's not something we can tune or optimise for - that's just broken hardware and the drive should quirk the brokeness away so nobody has to care about it. Tell the hardware manufacturer to fix their hardware, don't try to hack around it in software and then expect the user to know how to tune for that broken hardware.

Quoted 3 lines

In that case, we still have
a room to utilize FUA for small-sized writes such as filesystem metadata writes, but avoid DIO w/ FUA for sequential write stream.

Strawman.

We don't use FUA for normal DIO writes - they only get used for O_DSYNC writes, in which case we either use FUA if the device supports it, or we do a normal write followed by a cache flush. If there are metadata updates that the O_DSYNC needs to also flush, we don't use FUA by let the fileystem issue a cache flush in the most optimal possible after the write completes.

Either way, using O_DSYNC DIO writes for streaming, sequential data is a really poor choice for an application to make. Normal DIO writes followed by fdatasync() to flush the metadata and caches once will be much faster and far more efficient than a metadata and cache flush after every single data write, FUA or not.

-Dave.

Signature · 2 lines
Dave Chinner
[email protected]
author maintainer
#5 lore raw
In reply to Eric Biggers
Quoted 10 lines · Eric Biggers

[+Cc linux-block for FUA, and linux-xfs for iomap]

Quoted 4 lines · Jaegeuk Kim

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

I think there's still some benefit to use FUA such as small chunk writes for checkpoint.

Quoted 24 lines
Quoted 24 lines
---
 Documentation/ABI/testing/sysfs-fs-f2fs | 7 +++++++
 fs/f2fs/data.c                          | 2 ++
 fs/f2fs/f2fs.h                          | 1 +
 fs/f2fs/sysfs.c                         | 2 ++
 4 files changed, 12 insertions(+)
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
index 9b583dd0298b..cd96b09d7182 100644
@@ -434,6 +434,7 @@ Date:		April 2020
 Contact:	"Daeho Jeong" <[email protected]>
 Description:	Give a way to change iostat_period time. 3secs by default.
 		The new iostat trace gives stats gap given the period.
+
 What:		/sys/fs/f2fs/<disk>/max_io_bytes
 Date:		December 2020
 Contact:	"Jaegeuk Kim" <[email protected]>
@@ -442,6 +443,12 @@ Description:	This gives a control to limit the bio size in f2fs.
 		whereas, if it has a certain bytes value, f2fs won't submit a
 		bio larger than that size.
 
Quoted 8 lines

Documentation/ABI/testing/sysfs-fs-f2fs Description: This gives a control to limit the bio size in f2fs. · line 450

Quoted 5 lines
Documentation/ABI/testing/sysfs-fs-f2fs +5 -0
+What:		/sys/fs/f2fs/<disk>/no_fua_dio
+Date:		May 2022
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	This gives a signal to iomap, which should not use FUA for
+		direct IOs. Default: 0.

iomap is an implementation detail, so it shouldn't be mentioned in UAPI documentation. UAPI documentation should describe user-visible behavior only.

Ok.

Quoted 10 lines
Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
+
 What:		/sys/fs/f2fs/<disk>/stat/sb_status
 Date:		December 2020
 Contact:	"Chao Yu" <[email protected]>
fs/f2fs/data.c +0 -0
index f5f2b7233982..23486486eab2 100644
@@ -4153,6 +4153,8 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
Quoted 8 lines

fs/f2fs/data.c static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length, · line 4157

Quoted 5 lines
fs/f2fs/data.c +2 -0
 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
 	    offset + length > i_size_read(inode))
 		iomap->flags |= IOMAP_F_DIRTY;
+	if (F2FS_I_SB(inode)->no_fua_dio)
+		iomap->flags |= IOMAP_F_DIRTY;

This is overloading the IOMAP_F_DIRTY flag to mean something other than dirty. Perhaps this flag needs to be renamed, or a new flag should be added?

I'm not sure it's acceptable to add another flag for f2fs only.

Quoted 12 lines

fs/f2fs/f2fs.h struct f2fs_sb_info { · line 1674

Quoted 10 lines
fs/f2fs/f2fs.h +1 -0
index e10838879538..c2400ea0080b 100644
@@ -1671,6 +1671,7 @@ struct f2fs_sb_info {
 	int dir_level;				/* directory level */
 	int readdir_ra;				/* readahead inode in readdir */
 	u64 max_io_bytes;			/* max io bytes to merge IOs */
+	int no_fua_dio;				/* avoid FUA in DIO */

Make this a bool?

Done.

Quoted 13 lines
Quoted 13 lines
fs/f2fs/sysfs.c +1 -0
index 4c50aedd5144..24d628ca92cc 100644
@@ -771,6 +771,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, iostat_period_ms, iostat_period_ms);
 #endif
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, readdir_ra, readdir_ra);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_io_bytes, max_io_bytes);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, no_fua_dio, no_fua_dio);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_super_block, extension_list, extension_list);
 #ifdef CONFIG_F2FS_FAULT_INJECTION
Quoted 8 lines

fs/f2fs/sysfs.c static struct attribute *f2fs_attrs[] = { · line 894

Quoted 5 lines
fs/f2fs/sysfs.c +1 -0
@@ -890,6 +891,7 @@ static struct attribute *f2fs_attrs[] = {
 #endif
 	ATTR_LIST(readdir_ra),
 	ATTR_LIST(max_io_bytes),
+	ATTR_LIST(no_fua_dio),

Where is it validated that only valid values (0 or 1) can be written to this file?

Added.

Quoted 2 lines

- Eric

#6 lore raw
Quoted 69 lines · Jaegeuk Kim
Quoted 10 lines · Eric Biggers

[+Cc linux-block for FUA, and linux-xfs for iomap]

Quoted 4 lines · Jaegeuk Kim

Some UFS storage gives slower performance on FUA than write+cache_flush. Let's give a way to manage it.

Should the driver even be saying that it has FUA support in this case? If the driver didn't claim FUA support, that would also solve this problem.

I think there's still some benefit to use FUA such as small chunk writes for checkpoint.

Quoted 24 lines
Quoted 24 lines
---
 Documentation/ABI/testing/sysfs-fs-f2fs | 7 +++++++
 fs/f2fs/data.c                          | 2 ++
 fs/f2fs/f2fs.h                          | 1 +
 fs/f2fs/sysfs.c                         | 2 ++
 4 files changed, 12 insertions(+)
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
index 9b583dd0298b..cd96b09d7182 100644
@@ -434,6 +434,7 @@ Date:		April 2020
 Contact:	"Daeho Jeong" <[email protected]>
 Description:	Give a way to change iostat_period time. 3secs by default.
 		The new iostat trace gives stats gap given the period.
+
 What:		/sys/fs/f2fs/<disk>/max_io_bytes
 Date:		December 2020
 Contact:	"Jaegeuk Kim" <[email protected]>
@@ -442,6 +443,12 @@ Description:	This gives a control to limit the bio size in f2fs.
 		whereas, if it has a certain bytes value, f2fs won't submit a
 		bio larger than that size.
 
Quoted 8 lines

Documentation/ABI/testing/sysfs-fs-f2fs Description: This gives a control to limit the bio size in f2fs. · line 450

Quoted 5 lines
Documentation/ABI/testing/sysfs-fs-f2fs +5 -0
+What:		/sys/fs/f2fs/<disk>/no_fua_dio
+Date:		May 2022
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	This gives a signal to iomap, which should not use FUA for
+		direct IOs. Default: 0.

iomap is an implementation detail, so it shouldn't be mentioned in UAPI documentation. UAPI documentation should describe user-visible behavior only.

Ok.

Quoted 10 lines
Quoted 10 lines
Documentation/ABI/testing/sysfs-fs-f2fs +1 -0
+
 What:		/sys/fs/f2fs/<disk>/stat/sb_status
 Date:		December 2020
 Contact:	"Chao Yu" <[email protected]>
fs/f2fs/data.c +0 -0
index f5f2b7233982..23486486eab2 100644
@@ -4153,6 +4153,8 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
Quoted 8 lines

fs/f2fs/data.c static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length, · line 4157

Quoted 5 lines
fs/f2fs/data.c +2 -0
 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
 	    offset + length > i_size_read(inode))
 		iomap->flags |= IOMAP_F_DIRTY;
+	if (F2FS_I_SB(inode)->no_fua_dio)
+		iomap->flags |= IOMAP_F_DIRTY;

This is overloading the IOMAP_F_DIRTY flag to mean something other than dirty. Perhaps this flag needs to be renamed, or a new flag should be added?

Quoted 1 line

I'm not sure it's acceptable to add another flag for f2fs only.

I think Al and willy have been throwing around patches to tell iomap_dio_rw or someone that the caller will handle cache flushes and that it shouldn't initiate them on its own; would that help here?

--D

Quoted 41 lines
Quoted 12 lines

fs/f2fs/f2fs.h struct f2fs_sb_info { · line 1674

Quoted 10 lines
fs/f2fs/f2fs.h +1 -0
index e10838879538..c2400ea0080b 100644
@@ -1671,6 +1671,7 @@ struct f2fs_sb_info {
 	int dir_level;				/* directory level */
 	int readdir_ra;				/* readahead inode in readdir */
 	u64 max_io_bytes;			/* max io bytes to merge IOs */
+	int no_fua_dio;				/* avoid FUA in DIO */

Make this a bool?

Done.

Quoted 13 lines
Quoted 13 lines
fs/f2fs/sysfs.c +1 -0
index 4c50aedd5144..24d628ca92cc 100644
@@ -771,6 +771,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, iostat_period_ms, iostat_period_ms);
 #endif
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, readdir_ra, readdir_ra);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_io_bytes, max_io_bytes);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, no_fua_dio, no_fua_dio);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_super_block, extension_list, extension_list);
 #ifdef CONFIG_F2FS_FAULT_INJECTION
Quoted 8 lines

fs/f2fs/sysfs.c static struct attribute *f2fs_attrs[] = { · line 894

Quoted 5 lines
fs/f2fs/sysfs.c +1 -0
@@ -890,6 +891,7 @@ static struct attribute *f2fs_attrs[] = {
 #endif
 	ATTR_LIST(readdir_ra),
 	ATTR_LIST(max_io_bytes),
+	ATTR_LIST(no_fua_dio),

Where is it validated that only valid values (0 or 1) can be written to this file?

Added.

Quoted 2 lines

- Eric

#7 lore raw
In reply to Jaegeuk Kim
Quoted 2 lines · Jaegeuk Kim

I think there's still some benefit to use FUA such as small chunk writes for checkpoint.

Did you measure if there is? Because some SSDs basically implemented FUA as an implied flush after the write, in which case it would not really help there either (but also not hurt).

But as the previous two maintainers already said - this needs quirking at the driver layer, not in the submitter.

author maintainer
#8 lore raw
Quoted 10 lines · Christoph Hellwig
Quoted 2 lines · Jaegeuk Kim

I think there's still some benefit to use FUA such as small chunk writes for checkpoint.

Did you measure if there is? Because some SSDs basically implemented FUA as an implied flush after the write, in which case it would not really help there either (but also not hurt).

But as the previous two maintainers already said - this needs quirking at the driver layer, not in the submitter.

Thanks, I indeed measured this using UFS, and it turned out cache_flush is better than FUA all the time like this. Hence, I posted a quirk [1].

Write(us/KB) 4 64 256 1024 2048
FUA 873.792 754.604 995.624 1011.67 1067.99
CACHE_FLUSH 824.703 712.98 800.307 1019.5 1037.37

[1] https://lore.kernel.org/linux-scsi/[email protected]/T/#u

author maintainer
#9 lore raw
In reply to Jaegeuk Kim

Some UFS storage supporting FUA gives slower DIO write bandwidth with FUA than write+cache_flush. But, in some small chunk writes, there's no reason to avoid FUA for shorter latency. Let's give a way to handle it by user.

---

Note that, this is a RFC, waiting for a better/right solution.

 Documentation/ABI/testing/sysfs-fs-f2fs | 7 +++++++
 fs/f2fs/data.c                          | 2 ++
 fs/f2fs/f2fs.h                          | 1 +
 fs/f2fs/sysfs.c                         | 9 +++++++++
 4 files changed, 19 insertions(+)
Documentation/ABI/testing/sysfs-fs-f2fs +7 -0
index 9b583dd0298b..8ca49f7d28ad 100644
@@ -434,6 +434,7 @@ Date:		April 2020
 Contact:	"Daeho Jeong" <[email protected]>
 Description:	Give a way to change iostat_period time. 3secs by default.
 		The new iostat trace gives stats gap given the period.
+
 What:		/sys/fs/f2fs/<disk>/max_io_bytes
 Date:		December 2020
 Contact:	"Jaegeuk Kim" <[email protected]>
@@ -442,6 +443,12 @@ Description:	This gives a control to limit the bio size in f2fs.
 		whereas, if it has a certain bytes value, f2fs won't submit a
 		bio larger than that size.
 
+What:		/sys/fs/f2fs/<disk>/no_fua_dio
+Date:		May 2022
+Contact:	"Jaegeuk Kim" <[email protected]>
+Description:	This contorls whether direct IOs attach FUA or not. Default is
+		using FUA.
+
 What:		/sys/fs/f2fs/<disk>/stat/sb_status
 Date:		December 2020
 Contact:	"Chao Yu" <[email protected]>
fs/f2fs/data.c +2 -0
index f5f2b7233982..23486486eab2 100644
@@ -4153,6 +4153,8 @@ static int f2fs_iomap_begin(struct inode *inode, loff_t offset, loff_t length,
 	if ((inode->i_state & I_DIRTY_DATASYNC) ||
 	    offset + length > i_size_read(inode))
 		iomap->flags |= IOMAP_F_DIRTY;
+	if (F2FS_I_SB(inode)->no_fua_dio)
+		iomap->flags |= IOMAP_F_DIRTY;
 
 	return 0;
 }
fs/f2fs/f2fs.h +1 -0
index e10838879538..4897ada1929b 100644
@@ -1671,6 +1671,7 @@ struct f2fs_sb_info {
 	int dir_level;				/* directory level */
 	int readdir_ra;				/* readahead inode in readdir */
 	u64 max_io_bytes;			/* max io bytes to merge IOs */
+	bool no_fua_dio;			/* don't add FUA in DIO write */
 
 	block_t user_block_count;		/* # of user blocks */
 	block_t total_valid_block_count;	/* # of valid blocks */
fs/f2fs/sysfs.c +9 -0
index 4c50aedd5144..199ba3e20ab0 100644
@@ -608,6 +608,13 @@ static ssize_t __sbi_store(struct f2fs_attr *a,
 		return count;
 	}
 
+	if (!strcmp(a->attr.name, "no_fua_dio")) {
+		if (t != 0 && t != 1)
+			return -EINVAL;
+		sbi->no_fua_dio = t;
+		return count;
+	}
+
 	*ui = (unsigned int)t;
 
 	return count;
@@ -771,6 +778,7 @@ F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, iostat_period_ms, iostat_period_ms);
 #endif
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, readdir_ra, readdir_ra);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, max_io_bytes, max_io_bytes);
+F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, no_fua_dio, no_fua_dio);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_sb_info, gc_pin_file_thresh, gc_pin_file_threshold);
 F2FS_RW_ATTR(F2FS_SBI, f2fs_super_block, extension_list, extension_list);
 #ifdef CONFIG_F2FS_FAULT_INJECTION
@@ -890,6 +898,7 @@ static struct attribute *f2fs_attrs[] = {
 #endif
 	ATTR_LIST(readdir_ra),
 	ATTR_LIST(max_io_bytes),
+	ATTR_LIST(no_fua_dio),
 	ATTR_LIST(gc_pin_file_thresh),
 	ATTR_LIST(extension_list),
 #ifdef CONFIG_F2FS_FAULT_INJECTION
Signature · 1 line
2.36.1.124.g0e6072fb45-goog