f2fs: fix 32-bit build
PATCH 1/1 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 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.
--- fs/f2fs/segment.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
fs/f2fs/segment.c +3 -1
@@ -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: "",
Signature · 4 lines
2.9.0 ------------------------------------------------------------------------------
Arnd,
Quoted 4 lines
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.
Quoted 20 lines
Fixes 792b84b74b54 ("f2fs: support multiple devices")Signed-off-by Arnd Bergmann <[email protected]>--- fs/f2fs/segment.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)fs/f2fs/segment.c +1
-0@@ -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) {
fs/f2fs/segment.c static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi, · line 701
Quoted 9 lines
fs/f2fs/segment.c +2
-1@@ -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.
Signature · 7 lines
Damien Le Moal, Ph.D. Sr Manager, System Software Research Group, Western Digital [email protected] Tel: (+81) 0466-98-3593 (Ext. 51-3593) 1 kirihara-cho, Fujisawa, Kanagawa, 252-0888 Japan www.wdc.com, www.hgst.com
Jaegeuk Kim
maintainerHi 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,
Quoted 1 line
From
b709ff532daedc42526a46cb26860127f4959e01Mon Sep 17 00:00:00 2001
From: Arnd Bergmann <[email protected]>
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.
--- fs/f2fs/segment.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
fs/f2fs/segment.c +2 -1
@@ -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: "",
Signature · 1 line
2.8.3
Quoted 3 lines · CET Jaegeuk Kim
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.
fs/f2fs/segment.c static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi, · line 697
Quoted 7 lines
fs/f2fs/segment.c +1
-1@@ -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
------------------------------------------------------------------------------
Jaegeuk Kim
maintainerQuoted 9 lines · Arnd Bergmann
Quoted 3 lines · CET Jaegeuk Kim
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,
Quoted 12 lines
fs/f2fs/segment.c static int __f2fs_issue_discard_zone(struct f2fs_sb_info *sbi, · line 697
Quoted 8 lines
fs/f2fs/segment.c +1
-1@@ -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