[PATCH] f2fs: fix 32-bit build

[PATCH] f2fs: fix 32-bit build

[Arnd Bergmann]

The addition of multiple-device support broke CONFIG_BLK_DEV_ZONED
on 32-bit machines because of a 64-bit division:

fs/f2fs/f2fs.o: In function `__issue_discard_async':
extent_cache.c:(.text.__issue_discard_async+0xd4): undefined reference to `__aeabi_uldivmod'

Unfortunately, the sector number is usually a 64-bit number, and
we probably can't guarantee that bdev_zone_size() returns a
power-of-two number, so we actually have to do the expensive 64-bit
operation to get the remainder.

Fixes: 792b84b74b54 ("f2fs: support multiple devices")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 fs/f2fs/segment.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index 634834e5a232..e4c5497aa172 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -685,6 +685,7 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
 {
 	sector_t nr_sects = SECTOR_FROM_BLOCK(blklen);
 	sector_t sector;
+	u32 rem;
 	int devi = 0;
 
 	if (sbi->s_ndevs) {
@@ -693,7 +694,8 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
 	}
 	sector = SECTOR_FROM_BLOCK(blkstart);
 
-	if (sector % bdev_zone_size(bdev) || nr_sects != bdev_zone_size(bdev)) {
+	div_u64_rem(sector, bdev_zone_size(bdev), &rem);
+	if (rem || nr_sects != bdev_zone_size(bdev)) {
 		f2fs_msg(sbi->sb, KERN_INFO,
 			"(%d) %s: Unaligned discard attempted (block %x + %x)",
 			devi, sbi->s_ndevs ? FDEV(devi).path: "",
-- 
2.9.0


------------------------------------------------------------------------------

Re: [PATCH] f2fs: fix 32-bit build

[Damien Le Moal]

Arnd,

> Unfortunately, the sector number is usually a 64-bit number, and
> we probably can't guarantee that bdev_zone_size() returns a
> power-of-two number, so we actually have to do the expensive 64-bit
> operation to get the remainder.

No, the zone size is guaranteed to be a power of 2. See sd_zbc.c in Jens
linux-block tree, branch for-4.10/block. So it is ok to do a
& (bdev_zone_size() - 1) for the remainder.

That was indeed my mistake in this patch. Modulo should not have been
used in the first place. My bad.

> 
> Fixes: 792b84b74b54 ("f2fs: support multiple devices")
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
>  fs/f2fs/segment.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 634834e5a232..e4c5497aa172 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -685,6 +685,7 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
>  {
>  	sector_t nr_sects = SECTOR_FROM_BLOCK(blklen);
>  	sector_t sector;
> +	u32 rem;
>  	int devi = 0;
>  
>  	if (sbi->s_ndevs) {
> @@ -693,7 +694,8 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
>  	}
>  	sector = SECTOR_FROM_BLOCK(blkstart);
>  
> -	if (sector % bdev_zone_size(bdev) || nr_sects != bdev_zone_size(bdev)) {
> +	div_u64_rem(sector, bdev_zone_size(bdev), &rem);
> +	if (rem || nr_sects != bdev_zone_size(bdev)) {
>  		f2fs_msg(sbi->sb, KERN_INFO,
>  			"(%d) %s: Unaligned discard attempted (block %x + %x)",
>  			devi, sbi->s_ndevs ? FDEV(devi).path: "",

Thanks !

Best regards.

-- 
Damien Le Moal, Ph.D.
Sr Manager, System Software Research Group,
Western Digital
Damien.LeMoal@hgst.com
Tel: (+81) 0466-98-3593 (Ext. 51-3593)
1 kirihara-cho, Fujisawa, Kanagawa, 252-0888 Japan
www.wdc.com, www.hgst.com

Re: [PATCH] f2fs: fix 32-bit build

[Jaegeuk Kim]

Hi Arnd,

As Damien mentioned, is this fine to go, if you don't mind?
I've confirmed that sd_zbc.c allows a power-of-two number only.

Thanks,

>From b709ff532daedc42526a46cb26860127f4959e01 Mon Sep 17 00:00:00 2001
From: Arnd Bergmann <arnd@arndb.de>
Date: Tue, 22 Nov 2016 15:20:16 +0100
Subject: [PATCH] f2fs: fix 32-bit build

The addition of multiple-device support broke CONFIG_BLK_DEV_ZONED
on 32-bit machines because of a 64-bit division:

fs/f2fs/f2fs.o: In function `__issue_discard_async':
extent_cache.c:(.text.__issue_discard_async+0xd4): undefined reference to `__aeabi_uldivmod'

Unfortunately, the sector number is usually a 64-bit number, and
we guarantee that bdev_zone_size() returns a power-of-two number.

Fixes: 792b84b74b54 ("f2fs: support multiple devices")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
---
 fs/f2fs/segment.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
index eaa0b40..d5141a0 100644
--- a/fs/f2fs/segment.c
+++ b/fs/f2fs/segment.c
@@ -693,7 +693,8 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
 	}
 	sector = SECTOR_FROM_BLOCK(blkstart);
 
-	if (sector % bdev_zone_size(bdev) || nr_sects != bdev_zone_size(bdev)) {
+	if (sector & (bdev_zone_size(bdev) - 1) ||
+				nr_sects != bdev_zone_size(bdev)) {
 		f2fs_msg(sbi->sb, KERN_INFO,
 			"(%d) %s: Unaligned discard attempted (block %x + %x)",
 			devi, sbi->s_ndevs ? FDEV(devi).path: "",
-- 
2.8.3

Re: [PATCH] f2fs: fix 32-bit build

[Arnd Bergmann]

On Wednesday, November 23, 2016 10:42:26 AM CET Jaegeuk Kim wrote:
> 
> Unfortunately, the sector number is usually a 64-bit number, and
> we guarantee that bdev_zone_size() returns a power-of-two number.

The sentence no longer makes sense. Maybe

Fortunately, bdev_zone_size() is guaranteed to return a power-of-two
number, so we can replace the % operator with a cheaper bit mask.

> @@ -693,7 +693,8 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
>         }
>         sector = SECTOR_FROM_BLOCK(blkstart);
>  
> -       if (sector % bdev_zone_size(bdev) || nr_sects != bdev_zone_size(bdev)) {
> +       if (sector & (bdev_zone_size(bdev) - 1) ||
> 

looks good, thanks for following up!

	Arnd

------------------------------------------------------------------------------

Re: [PATCH] f2fs: fix 32-bit build

[Jaegeuk Kim]

On Wed, Nov 23, 2016 at 10:05:41PM +0100, Arnd Bergmann wrote:
> On Wednesday, November 23, 2016 10:42:26 AM CET Jaegeuk Kim wrote:
> > 
> > Unfortunately, the sector number is usually a 64-bit number, and
> > we guarantee that bdev_zone_size() returns a power-of-two number.
> 
> The sentence no longer makes sense. Maybe
> 
> Fortunately, bdev_zone_size() is guaranteed to return a power-of-two
> number, so we can replace the % operator with a cheaper bit mask.

Got it.
Added it with modifed version.
Thanks,

> 
> > @@ -693,7 +693,8 @@ static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi,
> >         }
> >         sector = SECTOR_FROM_BLOCK(blkstart);
> >  
> > -       if (sector % bdev_zone_size(bdev) || nr_sects != bdev_zone_size(bdev)) {
> > +       if (sector & (bdev_zone_size(bdev) - 1) ||
> > 
> 
> looks good, thanks for following up!
> 
> 	Arnd